Skip to contentSkip to conversation

sub-fetches discard --ipv4|6 option

PATCH In review Alex Riesen 3 participants 49 messages Sep 14, 2020 – Jan 7, 2021 Documentation

lore
Share & tools
lore
#0 lore raw

Hi everyone,

I have a slight problem with IPv6 configuration in my local network (connect works but transfers do not) and would like to temporarily disable use of the transport for a series of fetches. The fetches are all done from within a script to which I can pass options for "git fetch" commands in its command-line. The options will be appended to the fetch commands, i.e.:

    git fetch <hard-coded-script-options> --ipv4

Unfortunately, it only worked for the fetches which didn't use --all or --multiple. After a light searching, I failed to find an explanation as to why --all|--multiple are handled so inconsistently with single remote fetches and added the options (similar to --force or --keep) to the argument list for sub-fetches:

builtin/fetch.c +4 -0
index 82ac4be8a5..5e06c07106 100644
@@ -1531,6 +1531,10 @@ static void add_options_to_argv(struct argv_array *argv)
 		argv_array_push(argv, "-v");
 	else if (verbosity < 0)
 		argv_array_push(argv, "-q");
+	if (family == TRANSPORT_FAMILY_IPV4)
+		argv_array_push(argv, "--ipv4");
+	else if (family == TRANSPORT_FAMILY_IPV6)
+		argv_array_push(argv, "--ipv6");
 
 }
 

Am I missing something obvious?

Regards,
Alex

#1 lore raw
Quoted 21 lines · Alex Riesen

Unfortunately, it only worked for the fetches which didn't use --all or --multiple. After a light searching, I failed to find an explanation as to why --all|--multiple are handled so inconsistently with single remote fetches and added the options (similar to --force or --keep) to the argument list for sub-fetches:

builtin/fetch.c +4 -0
index 82ac4be8a5..5e06c07106 100644
@@ -1531,6 +1531,10 @@ static void add_options_to_argv(struct argv_array *argv)
 		argv_array_push(argv, "-v");
 	else if (verbosity < 0)
 		argv_array_push(argv, "-q");
+	if (family == TRANSPORT_FAMILY_IPV4)
+		argv_array_push(argv, "--ipv4");
+	else if (family == TRANSPORT_FAMILY_IPV6)
+		argv_array_push(argv, "--ipv6");
 
 }
 
Quoted 1 line

Am I missing something obvious?

I don't think so. When we're starting fetch sub-processes, some options will make sense to pass along and some won't. The parent has to either pass all options and omit some, or explicitly pass ones it knows are useful. It looks like the code chooses the latter, but these particular options never got added (and it seems like they should be, as they are only useful to the child fetch processes that actually touch the network).

So your patch above looks quite sensible (modulo useful bits like a signoff and maybe a test, though I guess the impact of those options is probably hard to cover in our tests).

It is rather unfortunate that anybody adding new fetch options needs to remember to (maybe) add them to add_options_to_argv() themselves.

Also, regarding these two specific options, it sounds like you'd want them set for all fetches during the time your IPv6 setup is broken. In which case I think a config option might have served you better. So that might be something worth implementing (though either way I think the fix above is worth doing independently).

-Peff

#2 lore raw

Jeff King, Mon, Sep 14, 2020 21:49:51 +0200:

Quoted 18 lines · Alex Riesen
Quoted 8 lines

Unfortunately, it only worked for the fetches which didn't use --all or --multiple. After a light searching, I failed to find an explanation as to why --all|--multiple are handled so inconsistently with single remote fetches and added the options (similar to --force or --keep) to the argument list for
sub-fetches: ...

Am I missing something obvious?

I don't think so. When we're starting fetch sub-processes, some options will make sense to pass along and some won't. The parent has to either pass all options and omit some, or explicitly pass ones it knows are useful. It looks like the code chooses the latter, but these particular options never got added (and it seems like they should be, as they are only useful to the child fetch processes that actually touch the network).

Quoted 3 lines

So your patch above looks quite sensible (modulo useful bits like a signoff and maybe a test, though I guess the impact of those options is probably hard to cover in our tests).

I tried to come up with one, but (aside from rather pointless checking of option presence in the trace output) failed to.

Or may be precisely this could be the point of the test: just do a fetch with all options we intend to pass down to sub-fetches and check that they are indeed present in the invocation of fetch --all/--multiple/--recurse-submodules?

Quoted 2 lines

It is rather unfortunate that anybody adding new fetch options needs to remember to (maybe) add them to add_options_to_argv() themselves.

Maybe make add_options_to_argv to go through builtin_fetch_options[] and copy the options with a special marker if they were provided?
And use the word "recursive" in help text as the marker :)

Quoted 5 lines

Also, regarding these two specific options, it sounds like you'd want them set for all fetches during the time your IPv6 setup is broken. In which case I think a config option might have served you better. So that might be something worth implementing (though either way I think the fix above is worth doing independently).

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

What would the preferred name for that config option be? fetch.ipv?

I'll be sending the first change reformatted as patch shortly. Just in case.

#3 lore raw

PATCH 1/1 Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

---

 builtin/fetch.c | 4 ++++
 1 file changed, 4 insertions(+)
builtin/fetch.c +4 -0
index 82ac4be8a5..447d28ac29 100644
@@ -1531,6 +1531,10 @@ static void add_options_to_argv(struct argv_array *argv)
 		argv_array_push(argv, "-v");
 	else if (verbosity < 0)
 		argv_array_push(argv, "-q");
+	if (family == TRANSPORT_FAMILY_IPV4)
+		argv_array_push(argv, "--ipv4");
+	else if (family == TRANSPORT_FAMILY_IPV6)
+		argv_array_push(argv, "--ipv6");
 
 }
 
Signature · 1 line
2.28.0.21.g09e033b31f
#4 lore raw

Re: [PATCH] Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

Quoted 7 lines · Alex Riesen

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

---

To avoid an overlong title and conform to project convention (aka "easier to read 'git shortlog --no-merges' output), I shortened the title and tweaked the text a bit to compensate for the change.

Thanks.

-- >8 --
From: Alex Riesen <[email protected]>
Date: Tue, 15 Sep 2020 13:54:07 +0200
Subject: [PATCH] fetch: pass --ipv4 and --ipv6 options to sub-fetches

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches (i.e. "--all" and recursive fetching of submodules) is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

---
 builtin/fetch.c | 4 ++++
 1 file changed, 4 insertions(+)
builtin/fetch.c +4 -0
index 82ac4be8a5..447d28ac29 100644
@@ -1531,6 +1531,10 @@ static void add_options_to_argv(struct argv_array *argv)
 		argv_array_push(argv, "-v");
 	else if (verbosity < 0)
 		argv_array_push(argv, "-q");
+	if (family == TRANSPORT_FAMILY_IPV4)
+		argv_array_push(argv, "--ipv4");
+	else if (family == TRANSPORT_FAMILY_IPV6)
+		argv_array_push(argv, "--ipv6");
 
 }
 
Signature · 1 line
2.28.0-618-g53f972bf7d
#5 lore raw

Re: [PATCH] Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

Junio C Hamano, Tue, Sep 15, 2020 23:19:11 +0200:

Quoted 11 lines
Quoted 5 lines · Alex Riesen

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

To avoid an overlong title and conform to project convention (aka "easier to read 'git shortlog --no-merges' output), I shortened the title and tweaked the text a bit to compensate for the change.

Thanks :)

#6 lore raw

Re: [PATCH] Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

In reply to Alex Riesen
Quoted 7 lines · Alex Riesen

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

---

Regardless of whether we move forward with the parse-options flag or config discussed in the other thread, I think this is an obvious improvement that we should take in the meantime.

-Peff

#7 lore raw

Re: [PATCH] Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

Quoted 7 lines · Jeff King
Quoted 5 lines · Alex Riesen

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

Quoted 6 lines
Quoted 2 lines

---

Regardless of whether we move forward with the parse-options flag or config discussed in the other thread, I think this is an obvious improvement that we should take in the meantime.

Yes. Others can wait. ipversion configuration variable is probably easier to sell; parse_options thing deserves a longer and deeper thought as it will affect the API future codebase would rely on.

Thanks.

#8 lore raw

Re: [PATCH] Pass --ipv4 and --ipv6 options to sub-fetches when fetching multiple remotes and submodules

Junio C Hamano, Wed, Sep 16, 2020 06:17:41 +0200:

Quoted 12 lines
Quoted 10 lines · Jeff King
Quoted 7 lines · Alex Riesen

The options indicate user intent for the whole fetch operation, and ignoring them in sub-fetches is quite unexpected when, for instance, it is intended to limit all of the communication to a specific transport protocol for some reason.

---

Quoted 6 lines
Quoted 3 lines

Regardless of whether we move forward with the parse-options flag or config discussed in the other thread, I think this is an obvious improvement that we should take in the meantime.

Yes. Others can wait. ipversion configuration variable is probably easier to sell; ...

Is its choice of namespace (transfer.) alright, too?

Quoted 2 lines

... parse_options thing deserves a longer and deeper thought as it will affect the API future codebase would rely on.

Of course.

Regards,
Alex

#9 lore raw
In reply to Alex Riesen
Quoted 10 lines · Alex Riesen
Quoted 3 lines

So your patch above looks quite sensible (modulo useful bits like a signoff and maybe a test, though I guess the impact of those options is probably hard to cover in our tests).

I tried to come up with one, but (aside from rather pointless checking of option presence in the trace output) failed to.

Or may be precisely this could be the point of the test: just do a fetch with all options we intend to pass down to sub-fetches and check that they are indeed present in the invocation of fetch --all/--multiple/--recurse-submodules?

Unfortunately I don't think that accomplishes much, since the main bug we're worried about is missing options. And it would require somebody adding the new options to the test, at which point you could just assume they would add it to add_options_to_argv().

Though I guess we can automatically get the list of options these days. So perhaps something like:

subopts=
for opt in $(git fetch --git-completion-helper)
do

        case "$opt" in
        # options that we know do not go to sub-fetches
        --all|--jobs|etc...)
                ;;
	# try/match only the positive versions
	--no-*)
	        ;;
	# give a fake value for options with values
	*=)
                subopts="$subopts ${opt}1"
		;;
	# and pass through any boolean options
	*)
                subopts="$subopts $opt"
		;;
        esac
  done
  GIT_TRACE=$PWD/trace.out git fetch --all $subopts
  perl -lne '
    BEGIN { @want = @ARGV; @ARGV = () }
    /run_command: git fetch (.*)/ and $seen{$_}++ for split(/ /, $1);
    END { print for grep { !$seen{$_} } @want }

' <trace.out -- $subopts

Except that doesn't quite work, because the parent fetch will complain about nonsense values (e.g., --filter=1). So it would probably need a bit more manual intelligence to cover those options. It looks like some options are mutually exclusive, too (--deepen/--depth), so maybe we'd need to run an individual "fetch --all" for each option.

I dunno. It's getting pretty complicated. :)

Quoted 6 lines
Quoted 2 lines

It is rather unfortunate that anybody adding new fetch options needs to remember to (maybe) add them to add_options_to_argv() themselves.

Maybe make add_options_to_argv to go through builtin_fetch_options[] and copy the options with a special marker if they were provided?
And use the word "recursive" in help text as the marker :)

Yeah, that would solve the duplication problem. We could probably add a "recursive" bit to the parse-options flag variable. Even if parse-options itself doesn't use it, it could be a convenience for callers like this one. It is a little inconvenient to set flags there, just because it usually means ditching our wrapper macros in favor of a raw struct declaration.

Quoted 6 lines

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

What would the preferred name for that config option be? fetch.ipv?

It looks like we've got similar options for clone/pull (which are really fetch under the hood of course) and push. We have the "transfer.*" namespace which applies to both already. So maybe "transfer.ipversion" or something?

-Peff

#10 lore raw

Jeff King, Tue, Sep 15, 2020 15:05:06 +0200:

Quoted 23 lines · Alex Riesen
Quoted 11 lines
Quoted 4 lines

So your patch above looks quite sensible (modulo useful bits like a signoff and maybe a test, though I guess the impact of those options is probably hard to cover in our tests).

I tried to come up with one, but (aside from rather pointless checking of option presence in the trace output) failed to.

Or may be precisely this could be the point of the test: just do a fetch with all options we intend to pass down to sub-fetches and check that they are indeed present in the invocation of fetch --all/--multiple/--recurse-submodules?

Unfortunately I don't think that accomplishes much, since the main bug we're worried about is missing options. And it would require somebody adding the new options to the test, at which point you could just assume they would add it to add_options_to_argv().

Though I guess we can automatically get the list of options these days. So perhaps something like:

  subopts=
  for opt in $(git fetch --git-completion-helper)

...

Quoted 7 lines

Except that doesn't quite work, because the parent fetch will complain about nonsense values (e.g., --filter=1). So it would probably need a bit more manual intelligence to cover those options. It looks like some options are mutually exclusive, too (--deepen/--depth), so maybe we'd need to run an individual "fetch --all" for each option.

I dunno. It's getting pretty complicated. :)

It does :-( And the manual parts will require perpetual maintenance. Not doing that yet than.

Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

It is rather unfortunate that anybody adding new fetch options needs to remember to (maybe) add them to add_options_to_argv() themselves.

Maybe make add_options_to_argv to go through builtin_fetch_options[] and copy the options with a special marker if they were provided?
And use the word "recursive" in help text as the marker :)

Quoted 6 lines

Yeah, that would solve the duplication problem. We could probably add a "recursive" bit to the parse-options flag variable. Even if parse-options itself doesn't use it, it could be a convenience for callers like this one. It is a little inconvenient to set flags there, just because it usually means ditching our wrapper macros in favor of a raw struct declaration.

Or extend the list of wrappers with _REC(URSIVE) macros

Regards,
Alex

#11 lore raw
Quoted 8 lines · Alex Riesen
Quoted 6 lines

Yeah, that would solve the duplication problem. We could probably add a "recursive" bit to the parse-options flag variable. Even if parse-options itself doesn't use it, it could be a convenience for callers like this one. It is a little inconvenient to set flags there, just because it usually means ditching our wrapper macros in favor of a raw struct declaration.

Or extend the list of wrappers with _REC(URSIVE) macros

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

  OPT_BOOL('f', "foo", &foo, "the foo option");

into:

  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

-Peff

#12 lore raw
Quoted 14 lines · Jeff King
Quoted 8 lines · Alex Riesen
Quoted 6 lines

Yeah, that would solve the duplication problem. We could probably add a "recursive" bit to the parse-options flag variable. Even if parse-options itself doesn't use it, it could be a convenience for callers like this one. It is a little inconvenient to set flags there, just because it usually means ditching our wrapper macros in favor of a raw struct declaration.

Or extend the list of wrappers with _REC(URSIVE) macros

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

Quoted 7 lines
  OPT_BOOL('f', "foo", &foo, "the foo option");

into:

  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

What is this "recursive" about? Does it have much in common with "passthru", or are they orthogonal?

#13 lore raw
Quoted 5 lines · Junio C Hamano
Quoted 5 lines

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

  OPT_BOOL('f', "foo", &foo, "the foo option");
Quoted 8 lines
Quoted 5 lines

into:

  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

What is this "recursive" about? Does it have much in common with "passthru", or are they orthogonal?

I agree the name is not super-descriptive. And it's a bit odd that the parse-options code itself would not care about it. It's simply a convenient bit for the calling code to use (rather than try to manage a separate array whose values correspond). It could be PARSE_OPT_USER1, but that is probably too inscrutable. :)

It's sort-of similar to passthru, but not quite. The problem with passthru is that it _always_ applies to the option. We never parse it as an option itself, but rather always stick its canonicalized form into an array of strings. But for the case we're talking about here, we don't know ahead of time whether we want the passthru behavior or not. It depends on whether we see an option like "--all" (which might even come after us).

So I think the best you could do is:

1. Keep two separate option lists, "parent" and "child". The parent list has "--all" in it. The child list has stuff like "--ipv6".

2. Parse using the parent list with PARSE_OPT_KEEP_UNKNOWN. That lets you decide whether we're in a mode that is spawning child fetch processes.

3. If we are spawning, then everything in the "child" option list becomes passthru. We could either mark them as such, or really, I guess we could just pass the remainder of argv on as-is (though it might be nice to diagnose a bogus config option once in the parent rather than in each child).

That would work OK. One downside is that PARSE_OPT_KEEP_UNKNOWN is inherently flaky. It doesn't know if "--foo --bar" is two options, or the option "--foo" with the value "--bar". You could solve that by leaving dummy passthru options in the "parent" list, but now we're back to having two copies.

I guess parse-options could provide a MAYBE_PASSTHRU flag. On the first parse_options() call, it would skip over any such options, leaving them in argv. On the second, the caller would tell it to actually parse them.

-Peff

#14 lore raw
Quoted 8 lines · Jeff King

So I think the best you could do is:

1. Keep two separate option lists, "parent" and "child". The parent list has "--all" in it. The child list has stuff like "--ipv6".

2. Parse using the parent list with PARSE_OPT_KEEP_UNKNOWN. That lets you decide whether we're in a mode that is spawning child fetch processes.

Hmph, I vaguely recall discussion about cascading options[] list but do not find anything that may be involved in an implementation like that in <parse-options.h>. I agree that neither of the above is so attractive.

Quoted 3 lines

I guess parse-options could provide a MAYBE_PASSTHRU flag. On the first parse_options() call, it would skip over any such options, leaving them in argv. On the second, the caller would tell it to actually parse them.

Or calling it USR1, which is a good way to make it crystal clear that parse_options() API does not do anything to it. The code like "builtin/fetch.c" can locally give it a more meaningful name with "#define PARSE_OPT_RECURSIVE PARSE_OPT_USR1". if recursive is the appropriate name for the bit in the context of the options[] array.

I agree that _F() convention that can be used across different types would be a good thing to have in the longer term, by the way.

Thanks.

#15 lore raw
Quoted 7 lines · Junio C Hamano
Quoted 5 lines · Jeff King

So I think the best you could do is:

1. Keep two separate option lists, "parent" and "child". The parent list has "--all" in it. The child list has stuff like "--ipv6".

Quoted 8 lines
Quoted 3 lines

2. Parse using the parent list with PARSE_OPT_KEEP_UNKNOWN. That lets you decide whether we're in a mode that is spawning child fetch processes.

Hmph, I vaguely recall discussion about cascading options[] list but do not find anything that may be involved in an implementation like that in <parse-options.h>. I agree that neither of the above is so attractive.

I think we just use KEEP_UNKNOWN in those cases and ignore any downsides to it.

Quoted 9 lines
Quoted 3 lines

I guess parse-options could provide a MAYBE_PASSTHRU flag. On the first parse_options() call, it would skip over any such options, leaving them in argv. On the second, the caller would tell it to actually parse them.

Or calling it USR1, which is a good way to make it crystal clear that parse_options() API does not do anything to it. The code like "builtin/fetch.c" can locally give it a more meaningful name with "#define PARSE_OPT_RECURSIVE PARSE_OPT_USR1". if recursive is the appropriate name for the bit in the context of the options[] array.

Ah, that's a good suggestion. My earlier "USER" suggestion was tongue-in-cheek, because I think it makes the resulting options list quite confusing. But a local #define fixes that nicely.

That said, it sounds from the other part of the thread like we'll need better parse-options support anyway, so this "noop flag bit" idea probably isn't a good direction anyway.

-Peff

#16 lore raw
Quoted 3 lines · Jeff King

That said, it sounds from the other part of the thread like we'll need better parse-options support anyway, so this "noop flag bit" idea probably isn't a good direction anyway.

OK. Thanks.

#17 lore raw
In reply to Jeff King

Jeff King, Tue, Sep 15, 2020 17:27:30 +0200:

Quoted 14 lines · Alex Riesen
Quoted 9 lines
Quoted 7 lines

Yeah, that would solve the duplication problem. We could probably add a "recursive" bit to the parse-options flag variable. Even if parse-options itself doesn't use it, it could be a convenience for callers like this one. It is a little inconvenient to set flags there, just because it usually means ditching our wrapper macros in favor of a raw struct declaration.

Or extend the list of wrappers with _REC(URSIVE) macros

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

Quoted 7 lines
  OPT_BOOL('f', "foo", &foo, "the foo option");

into:

  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

This part (marking of the options) was easy. What's left is finding out if an option was actually specified in the command-line. The ...options[] arrays are not update by parse_options() with what was given, are they?

Maybe extend struct option with a field to store given command-line argument (as it was specified) and parse_options() will update the field if PARSE_OPT_RECURSIVE is present in .flags?
Is it allowed for parse_options() to modify the options array? Or is it possible to use something in parse-options.h API to note the arguments somewhere while they are parse? I mean, there are parse_options_start/step/end, can cmd_fetch argument parsing use those so that the options marked recursive can be saved for sub-fetches?

#18 lore raw
Quoted 7 lines · Alex Riesen
Quoted 7 lines

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

  OPT_BOOL('f', "foo", &foo, "the foo option");

into:

Quoted 7 lines
Quoted 3 lines
  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

This part (marking of the options) was easy. What's left is finding out if an option was actually specified in the command-line. The ...options[] arrays are not update by parse_options() with what was given, are they?

Oh right. Having the list of options is not that helpful because add_options_argv() is actually working off the parsed data in individual variables. Sorry for leading you in a (maybe) wrong direction.

I think this approach would have to be coupled with some mechanism for looking over the original list of options (either saving the original argv before parsing, or teaching parse-options the kind of two-pass "don't parse these the first time" mechanism discussed elsewhere in the thread).

Quoted 4 lines

Maybe extend struct option with a field to store given command-line argument (as it was specified) and parse_options() will update the field if PARSE_OPT_RECURSIVE is present in .flags?
Is it allowed for parse_options() to modify the options array?

No, we take the options array as a const pointer, so many callers would likely need to be updated to handle that. Plus it's possible some may actually re-use the array multiple times in some cases.

Quoted 4 lines

Or is it possible to use something in parse-options.h API to note the arguments somewhere while they are parse? I mean, there are parse_options_start/step/end, can cmd_fetch argument parsing use those so that the options marked recursive can be saved for sub-fetches?

Possibly the step-wise parsing could help. But I think it might be easier to just let parse_options() save a copy of parsed options. And then our PARSE_OPT_RECURSIVE really becomes PARSE_OPT_SAVE or similar, which would cause parse-options to save the original option (and any value argument) in its original form.

There's one slight complication, which is how the array of saved options gets communicated back to the caller. Leaving them in the original argv probably isn't a good idea (because the caller relies on it having options removed in order to find the non-option arguments).

Adding a new strvec pointer to parse_options() works, but means updating all of the callers, most of which will pass NULL. Possibly the existing "flags" parameter to parse_options() could grow into a struct. That requires modifying each caller, but at least solves the problem once and for all.

Another option is to stick it into parse_opt_ctx_t. That's used only be step-wise callers, of which there are very few.

-Peff

#19 lore raw

Jeff King, Wed, Sep 16, 2020 18:32:18 +0200:

Quoted 20 lines · Alex Riesen
Quoted 8 lines
Quoted 8 lines

If you go that route, we have some "_F" macros that take flags. Probably would make sense to add it more consistently, which lets you convert:

  OPT_BOOL('f', "foo", &foo, "the foo option");

into:

Quoted 7 lines
Quoted 3 lines
  OPT_BOOL_F('f', "foo", &foo, "the foo option", PARSE_OPT_RECURSIVE);

but could also be used for other flags.

This part (marking of the options) was easy. What's left is finding out if an option was actually specified in the command-line. The ...options[] arrays are not update by parse_options() with what was given, are they?

Oh right. Having the list of options is not that helpful because add_options_argv() is actually working off the parsed data in individual variables. Sorry for leading you in a (maybe) wrong direction.

...

Quoted 16 lines
Quoted 4 lines

Or is it possible to use something in parse-options.h API to note the arguments somewhere while they are parsed? I mean, there are parse_options_start/step/end, can cmd_fetch argument parsing use those so that the options marked recursive can be saved for sub-fetches?

Possibly the step-wise parsing could help. But I think it might be easier to just let parse_options() save a copy of parsed options. And then our PARSE_OPT_RECURSIVE really becomes PARSE_OPT_SAVE or similar, which would cause parse-options to save the original option (and any value argument) in its original form.

There's one slight complication, which is how the array of saved options gets communicated back to the caller. Leaving them in the original argv probably isn't a good idea (because the caller relies on it having options removed in order to find the non-option arguments).

Quoted 5 lines

Adding a new strvec pointer to parse_options() works, but means updating all of the callers, most of which will pass NULL. Possibly the existing "flags" parameter to parse_options() could grow into a struct. That requires modifying each caller, but at least solves the problem once and for all.

With such complication a step-wise parsing sounds easier, given that at the moment there is only one user for the feature. Are there *existing* callers of parse_options with similar requirements?

I feel that doing this kind of selection work in parse_options is an overkill:
if it is specific for just this use case, the implementation might be more complex than necessary, while profiting just one caller.

Quoted 2 lines

Another option is to stick it into parse_opt_ctx_t. That's used only be step-wise callers, of which there are very few.

Does that mean that currently there is no way to find out which option corresponds to the last parsed command-line argument after a call to parse_options_step? Which in turn makes the marking of recursive options inaccessible to step-wise command line parsing code, right?

#20 lore raw
Quoted 9 lines · Alex Riesen
Quoted 5 lines

Adding a new strvec pointer to parse_options() works, but means updating all of the callers, most of which will pass NULL. Possibly the existing "flags" parameter to parse_options() could grow into a struct. That requires modifying each caller, but at least solves the problem once and for all.

With such complication a step-wise parsing sounds easier, given that at the moment there is only one user for the feature. Are there *existing* callers of parse_options with similar requirements?

I don't know offhand. I'd suspect some of the command which take --recurse-submodules do something similar, but the number of steps between the main command the submodule argv may make it awkward to use a parse-options solution. I don't think we have a "push to each of these remotes" option the way we do for fetch.

Quoted 3 lines

I feel that doing this kind of selection work in parse_options is an overkill:
if it is specific for just this use case, the implementation might be more complex than necessary, while profiting just one caller.

Yeah, I agree it's complex, and I'm happy with simpler solutions (or just fixing these ones as you did and punting on it for now).

Quoted 7 lines
Quoted 2 lines

Another option is to stick it into parse_opt_ctx_t. That's used only be step-wise callers, of which there are very few.

Does that mean that currently there is no way to find out which option corresponds to the last parsed command-line argument after a call to parse_options_step? Which in turn makes the marking of recursive options inaccessible to step-wise command line parsing code, right?

I'm not super familiar with the internals of parse-options, but it doesn't look like it. Each step consumes an argv and matches it to a "struct option", but I don't think you get to know which struct it was matched to. It would be reasonable for it to keep a pointer in the parse_opt_ctx_t (and of course you'd need some bit in the option struct itself to say "I am a recursive option").

-Peff

#21 lore raw

PATCH 1/1 config: option transfer.ipversion to set transport protocol version for network fetches

In reply to Jeff King

Affecting the transfers caused by git-fetch, the option allows to control network operations similar to --ipv4 and --ipv6 options.

---

Jeff King, Tue, Sep 15, 2020 15:05:06 +0200:

Quoted 6 lines · Alex Riesen
Quoted 5 lines

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

Quoted 6 lines
Quoted 1 line

What would the preferred name for that config option be? fetch.ipv?

It looks like we've got similar options for clone/pull (which are really fetch under the hood of course) and push. We have the "transfer.*" namespace which applies to both already. So maybe "transfer.ipversion" or something?

Something like this?

 Documentation/config/transfer.txt |  7 +++++++
 builtin/fetch.c                   | 11 +++++++++++
 2 files changed, 18 insertions(+)
Documentation/config/transfer.txt +7 -0
index f5b6245270..cc0e97fbb1 100644
@@ -69,3 +69,10 @@ transfer.unpackLimit::
 	When `fetch.unpackLimit` or `receive.unpackLimit` are
 	not set, the value of this variable is used instead.
 	The default value is 100.
+
+transfer.ipversion::
+	Limit the network operations to the specified version of the transport
+	protocol. Can be specified as `4` to allow IPv4 only, `6` for IPv6, or
+	`all` to allow all protocols.
+	See also linkgit:git-fetch[1] options `--ipv4` and `--ipv6`.
+	The default value is `all` to allow all protocols.
builtin/fetch.c +11 -0
index 447d28ac29..da01c8f7b3 100644
@@ -118,6 +118,17 @@ static int git_fetch_config(const char *k, const char *v, void *cb)
 		return 0;
 	}
 
+	if (!strcmp(k, "transfer.ipversion")) {
+		if (!strcmp(v, "all"))
+			;
+		else if (!strcmp(v, "4"))
+			family = TRANSPORT_FAMILY_IPV4;
+		else if (!strcmp(v, "6"))
+			family = TRANSPORT_FAMILY_IPV6;
+		else
+			die(_("transfer.ipversion can be only 4, 6, or any"));
+		return 0;
+	}
 	return git_default_config(k, v, cb);
 }
 
Signature · 1 line
2.28.0.21.g178b32a6fd.dirty
#22 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 7 lines · Alex Riesen

Jeff King, Tue, Sep 15, 2020 15:05:06 +0200:

Quoted 6 lines · Alex Riesen
Quoted 5 lines

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

What would the preferred name for that config option be? fetch.ipv?

It looks like we've got similar options for clone/pull (which are really fetch under the hood of course) and push. We have the "transfer.*" namespace which applies to both already. So maybe "transfer.ipversion" or something?

Something like this?

That's the right direction, but I think we'd want to make sure it impacted all of the spots that allow switching. "clone" on the fetching side, but probably also "push".

Each of those commands could learn the same config, but it might be easier to just enforce it in the transport layer. Something like this maybe (compiled but not tested):

builtin/fetch.c +2 -1
index a6d3268661..2f7a734eb2 100644
@@ -1267,7 +1267,8 @@ static struct transport *prepare_transport(struct remote *remote, int deepen)
 
 	transport = transport_get(remote, NULL);
 	transport_set_verbosity(transport, verbosity, progress);
-	transport->family = family;
+	if (family)
+		transport->family = family;
 	if (upload_pack)
 		set_option(transport, TRANS_OPT_UPLOADPACK, upload_pack);
 	if (keep)
builtin/push.c +2 -1
index bc94078e72..f7a40b65cd 100644
@@ -343,7 +343,8 @@ static int push_with_options(struct transport *transport, struct refspec *rs,
 	char *anon_url = transport_anonymize_url(transport->url);
 
 	transport_set_verbosity(transport, verbosity, progress);
-	transport->family = family;
+	if (family)
+		transport->family = family;
 
 	if (receivepack)
 		transport_set_option(transport,
transport.c +16 -0
index 43e24bf1e5..92f81b414d 100644
@@ -919,6 +919,20 @@ static struct transport_vtable builtin_smart_vtable = {
 	disconnect_git
 };
 
+enum transport_family default_transport_family(void)
+{
+	const char *v;
+
+	if (git_config_get_string_tmp("core.ipversion", &v))
+		return TRANSPORT_FAMILY_ALL;
+	else if (!strcmp(v, "ipv4"))
+		return TRANSPORT_FAMILY_IPV4;
+	else if (!strcmp(v, "ipv6"))
+		return TRANSPORT_FAMILY_IPV6;
+
+	die(_("invalid core.ipversion: %s"), v);
+}
+
 struct transport *transport_get(struct remote *remote, const char *url)
 {
 	const char *helper;
@@ -948,6 +962,8 @@ struct transport *transport_get(struct remote *remote, const char *url)
 			helper = xstrndup(url, p - url);
 	}
 
+	ret->family = default_transport_family();
+
 	if (helper) {
 		transport_helper_init(ret, helper);
 	} else if (starts_with(url, "rsync:")) {

A few notes:

- it cheats a little by noting that command-line options can't get it to "ALL". It might be a bit cleaner if we have an explicit TRANSPORT_FAMILY_UNSET, and then resolve the default at look-up time.

- it probably needs more "if (family)" spots in each command, which is unfortunate (which again might go away if "UNSET" is 0).

- I waffled on the name. transfer.* is where we put things that apply to both push/fetch, but it's usually more related to the git layer. This is more of a core networking decision, and if we added other network programs, I'd expect them to respect it. So maybe "core." is a better namespace.

-Peff

#23 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Jeff King, Wed, Sep 16, 2020 22:02:03 +0200:

Quoted 16 lines · Alex Riesen
Quoted 15 lines

Jeff King, Tue, Sep 15, 2020 15:05:06 +0200:

Quoted 12 lines · Alex Riesen
Quoted 6 lines

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

What would the preferred name for that config option be? fetch.ipv?

It looks like we've got similar options for clone/pull (which are really fetch under the hood of course) and push. We have the "transfer.*" namespace which applies to both already. So maybe "transfer.ipversion" or something?

Quoted 5 lines
Quoted 1 line

Something like this?

That's the right direction, but I think we'd want to make sure it impacted all of the spots that allow switching. "clone" on the fetching side, but probably also "push".

Ah sorry. Missed that push case.

#24 lore raw

PATCH 1/1 Config option to set the transport protocol version for network fetches

In reply to Jeff King

Affecting the transfers initiated by fetch and push, the option allows to control network operations similar to --ipv4 and --ipv6 options.

---

Jeff King, Wed, Sep 16, 2020 22:02:03 +0200:

Quoted 18 lines · Alex Riesen
Quoted 8 lines

Jeff King, Tue, Sep 15, 2020 15:05:06 +0200:

Quoted 6 lines · Alex Riesen
Quoted 5 lines

Sure! Thinking about it, I actually would have preferred to have both: a config option and a command-line option. So that I can set --ipv4 in, say, ~/.config/git/config file, but still have the option to try --ipv6 from time to time to check if the network setup magically fixed itself.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

What would the preferred name for that config option be? fetch.ipv?

It looks like we've got similar options for clone/pull (which are really fetch under the hood of course) and push. We have the "transfer.*" namespace which applies to both already. So maybe "transfer.ipversion" or something?

Something like this?

Quoted 7 lines

That's the right direction, but I think we'd want to make sure it impacted all of the spots that allow switching. "clone" on the fetching side, but probably also "push".

Each of those commands could learn the same config, but it might be easier to just enforce it in the transport layer. Something like this maybe (compiled but not tested):

I have merged the patches.

Quoted 9 lines

A few notes:

- it cheats a little by noting that command-line options can't get it to "ALL". It might be a bit cleaner if we have an explicit TRANSPORT_FAMILY_UNSET, and then resolve the default at look-up time.

- it probably needs more "if (family)" spots in each command, which is unfortunate (which again might go away if "UNSET" is 0).

As this is still being discussed, and because I still imagine the transport protocol family as a list of allowed protocols, this part is not in the patch below.

Quoted 5 lines

- I waffled on the name. transfer.* is where we put things that apply to both push/fetch, but it's usually more related to the git layer. This is more of a core networking decision, and if we added other network programs, I'd expect them to respect it. So maybe "core." is a better namespace.

I agree.

 Documentation/config/core.txt |  7 +++++++
 builtin/fetch.c               |  3 ++-
 builtin/push.c                |  3 ++-
 transport.c                   | 19 +++++++++++++++++++
 4 files changed, 30 insertions(+), 2 deletions(-)
Documentation/config/core.txt +7 -0
index 74619a9c03..dcb7db9799 100644
@@ -626,3 +626,10 @@ core.abbrev::
 	in your repository, which hopefully is enough for
 	abbreviated object names to stay unique for some time.
 	The minimum length is 4.
+
+core.ipversion::
+	Limit the network operations to the specified version of the transport
+	protocol. Can be specified as `4` to allow IPv4 only, `6` for IPv6, or
+	`all` to allow all protocols.
+	See also linkgit:git-fetch[1] options `--ipv4` and `--ipv6`.
+	The default value is `all` to allow all protocols.
builtin/fetch.c +2 -1
index 447d28ac29..41f82d61d7 100644
@@ -1248,7 +1248,8 @@ static struct transport *prepare_transport(struct remote *remote, int deepen)
 
 	transport = transport_get(remote, NULL);
 	transport_set_verbosity(transport, verbosity, progress);
-	transport->family = family;
+	if (family)
+		transport->family = family;
 	if (upload_pack)
 		set_option(transport, TRANS_OPT_UPLOADPACK, upload_pack);
 	if (keep)
builtin/push.c +2 -1
index bc94078e72..f7a40b65cd 100644
@@ -343,7 +343,8 @@ static int push_with_options(struct transport *transport, struct refspec *rs,
 	char *anon_url = transport_anonymize_url(transport->url);
 
 	transport_set_verbosity(transport, verbosity, progress);
-	transport->family = family;
+	if (family)
+		transport->family = family;
 
 	if (receivepack)
 		transport_set_option(transport,
transport.c +19 -0
index b41386eccb..e16c339f3e 100644
@@ -922,6 +922,23 @@ static struct transport_vtable builtin_smart_vtable = {
 	disconnect_git
 };
 
+static enum transport_family default_transport_family(void)
+{
+	static const char key[] = "core.ipversion";
+	const char *v;
+
+	if (git_config_get_string_const(key, &v))
+		return TRANSPORT_FAMILY_ALL;
+	if (!strcmp(v, "all"))
+		return TRANSPORT_FAMILY_ALL;
+	if (!strcmp(v, "ipv4"))
+		return TRANSPORT_FAMILY_IPV4;
+	if (!strcmp(v, "ipv6"))
+		return TRANSPORT_FAMILY_IPV6;
+
+	die(_("%s: unknown value '%s'"), key, v);
+}
+
 struct transport *transport_get(struct remote *remote, const char *url)
 {
 	const char *helper;
@@ -951,6 +968,8 @@ struct transport *transport_get(struct remote *remote, const char *url)
 			helper = xstrndup(url, p - url);
 	}
 
+	ret->family = default_transport_family();
+
 	if (helper) {
 		transport_helper_init(ret, helper);
 	} else if (starts_with(url, "rsync:")) {
Signature · 1 line
2.28.0.22.gfab0d4627e
#25 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Quoted 6 lines · Alex Riesen

Affecting the transfers initiated by fetch and push, the option allows to control network operations similar to --ipv4 and --ipv6 options.

I think this misses some of the excellent suggestions from Junio (naming, and the ability to override from the command line).

-Peff

#26 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Jeff King, Thu, Sep 17, 2020 15:31:53 +0200:

Quoted 11 lines · Alex Riesen
Quoted 7 lines

Affecting the transfers initiated by fetch and push, the option allows to control network operations similar to --ipv4 and --ipv6 options.

I think this misses some of the excellent suggestions from Junio (naming, and the ability to override from the command line).

It does, sorry. Also the suggestions to the issue of consistently passing the options to helper programs haven't been collected.
Haven't had the time yet.

#27 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Quoted 10 lines · Alex Riesen

Jeff King, Thu, Sep 17, 2020 15:31:53 +0200:

Quoted 9 lines · Alex Riesen
Quoted 7 lines

Affecting the transfers initiated by fetch and push, the option allows to control network operations similar to --ipv4 and --ipv6 options.

Quoted 6 lines
Quoted 2 lines

I think this misses some of the excellent suggestions from Junio (naming, and the ability to override from the command line).

It does, sorry. Also the suggestions to the issue of consistently passing the options to helper programs haven't been collected.
Haven't had the time yet.

No problem, and no rush. I just wanted to make sure those bits didn't get overlooked.

-Peff

#28 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Jeff King, Thu, Sep 17, 2020 16:51:42 +0200:

Quoted 2 lines

No problem, and no rush. I just wanted to make sure those bits didn't get overlooked.

I'll try and do my best :)

#29 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Quoted 5 lines · Alex Riesen

Jeff King, Thu, Sep 17, 2020 16:51:42 +0200:

Quoted 2 lines

No problem, and no rush. I just wanted to make sure those bits didn't get overlooked.

I'll try and do my best :)

If I remember correctly, this discussion started as an introduction of useful feature, but got stuck at the implementation phase of how the feature is presented at the UI level after we had general concensus on the design.

It has been about 3 months; has anything happened that I missed? It seems that I kept the thread-starter patch in the 'seen' branch, but and haven't updated with a later attempt in the discussion.

Thanks.

#30 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

Junio C Hamano, Tue, Dec 22, 2020 20:55:25 +0100:

Quoted 5 lines
Quoted 4 lines · Alex Riesen

Jeff King, Thu, Sep 17, 2020 16:51:42 +0200:

Quoted 2 lines

No problem, and no rush. I just wanted to make sure those bits didn't get overlooked.

Quoted 8 lines
Quoted 1 line

I'll try and do my best :)

If I remember correctly, this discussion started as an introduction of useful feature, but got stuck at the implementation phase of how the feature is presented at the UI level after we had general concensus on the design.

It has been about 3 months; has anything happened that I missed?

No, nothing happened.

Quoted 2 lines

It seems that I kept the thread-starter patch in the 'seen' branch, but and haven't updated with a later attempt in the discussion.

Sorry. I got very distracted. As my situation seems to persist, I cannot promise to do something about it soon.

I keep the branch and the discussion close by, though. Just in case.

#31 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

In reply to Alex Riesen

Alex Riesen, Thu, Sep 17, 2020 15:20:47 +0200:

Documentation/config/core.txt core.abbrev:: · line 633

Quoted 13 lines
Documentation/config/core.txt +5 -0
index 74619a9c03..dcb7db9799 100644
@@ -626,3 +626,10 @@ core.abbrev::
 	in your repository, which hopefully is enough for
 	abbreviated object names to stay unique for some time.
 	The minimum length is 4.
+
+core.ipversion::
+	Limit the network operations to the specified version of the transport
+	protocol. Can be specified as `4` to allow IPv4 only, `6` for IPv6, or
+	`all` to allow all protocols.

Eh. Option values are "ipv4" and "ipv6" indeed, not "4" and "6".

And I compiled and ran the code by now. Feels ok.

#32 lore raw

Re: [PATCH] Config option to set the transport protocol version for network fetches

In reply to Alex Riesen

Alex Riesen, Thu, Sep 17, 2020 15:20:47 +0200:

Quoted 5 lines
transport.c +0 -0
index b41386eccb..e16c339f3e 100644

transport.c static struct transport_vtable builtin_smart_vtable = { · line 930

Quoted 9 lines
transport.c +6 -0
@@ -922,6 +922,23 @@ static struct transport_vtable builtin_smart_vtable = {
 	disconnect_git
 };
 
+static enum transport_family default_transport_family(void)
+{
+	static const char key[] = "core.ipversion";
+	const char *v;
+
+	if (git_config_get_string_const(key, &v))

Sorry about that. git_config_get_string_tmp, indeed.

#33 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

In reply to Alex Riesen
Quoted 5 lines · Alex Riesen

Affecting the transfers caused by git-fetch, the option allows to control network operations similar to --ipv4 and --ipv6 options.
...
Something like this?

Good start.

If I configure it to "4", I do not see a way to override it and say "I don't care which one is used". With the introduction of this configuration variable, we'd need a bit more support on the command line side.

How about introducing a new command line option

	--transfer-protocol-family=("any"|<protocol>)

where <protocol> is either "ipv4" or "ipv6" [*1*], and make existing "--ipv4" a synonym for "--transfer-protocol-family=ipv4" and "--ipv6" for "--transfer-protocol-family=ipv6"

With such an extended command line override, we can override configured

[transfer]
ipversion = 6

with "--transfer-protocol-family=any" from the command line.

Also, we should follow the usual "the last one wins" for a configuration variable like this, which is *not* a multi-valued variable. So the config parsing would look more like this:

	if (!strcmp(k, "transfer.ipversion")) {
		if (!v)
			return config_error_nonbool("transfer.ipversion");
		if (!strcmp(v, "any"))
			family = 0;
		else if (!strcmp(v, "4") || !strcmp(v, "ipv4"))
			family = TRANSPORT_FAMILY_IPV4;
		else if (!strcmp(v, "6") || !strcmp(v, "ipv6"))
			family = TRANSPORT_FAMILY_IPV6;
		else
			return error("transfer.ipversion: unknown value '%s'", v);
	}

Would we regret to choose 'ipversion' as the variable name, by the way? On the command line side, --transfer-protocol-family=ipv4 makes it clear that we leave room to support protocols outside the Internet protocol family, and existing --ipv4 is grandfathered in by making it a synonym to --transfer-protocol-family=ipv4. Calling the variable "transfer.ipversion" and still allowing future protocols outside the Internet protocol family is rather awkward.

Calling "transfer.protocolFamily" would not have such a problem, though.

[Footnote]

*1* But leave a room to extend it in the future to a comma-separated
list of them to allow something like "ipv6,ipv7,ipv8" (i.e. "not just 'any'---we want to say that 'ipv4' is not welcomed").

Quoted 40 lines
 Documentation/config/transfer.txt |  7 +++++++
 builtin/fetch.c                   | 11 +++++++++++
 2 files changed, 18 insertions(+)
Documentation/config/transfer.txt +7 -0
index f5b6245270..cc0e97fbb1 100644
@@ -69,3 +69,10 @@ transfer.unpackLimit::
 	When `fetch.unpackLimit` or `receive.unpackLimit` are
 	not set, the value of this variable is used instead.
 	The default value is 100.
+
+transfer.ipversion::
+	Limit the network operations to the specified version of the transport
+	protocol. Can be specified as `4` to allow IPv4 only, `6` for IPv6, or
+	`all` to allow all protocols.
+	See also linkgit:git-fetch[1] options `--ipv4` and `--ipv6`.
+	The default value is `all` to allow all protocols.
builtin/fetch.c +11 -0
index 447d28ac29..da01c8f7b3 100644
@@ -118,6 +118,17 @@ static int git_fetch_config(const char *k, const char *v, void *cb)
 		return 0;
 	}
 
+	if (!strcmp(k, "transfer.ipversion")) {
+		if (!strcmp(v, "all"))
+			;
+		else if (!strcmp(v, "4"))
+			family = TRANSPORT_FAMILY_IPV4;
+		else if (!strcmp(v, "6"))
+			family = TRANSPORT_FAMILY_IPV6;
+		else
+			die(_("transfer.ipversion can be only 4, 6, or any"));
+		return 0;
+	}
 	return git_default_config(k, v, cb);
 }
#34 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 10 lines · Junio C Hamano
Quoted 5 lines · Alex Riesen

Affecting the transfers caused by git-fetch, the option allows to control network operations similar to --ipv4 and --ipv6 options.
...
Something like this?

Good start.
[...]

A lot of your comments apply to the "something like this" suggestion I just posted, so I wanted to save a round-trip and say: yes, I agree with all of your suggestions here.

Adding a command-line option for "all" is a good idea, but will probably mean needing to add the "unset" sentinel value I mentioned in the other email.

Quoted 10 lines

Would we regret to choose 'ipversion' as the variable name, by the way? On the command line side, --transfer-protocol-family=ipv4 makes it clear that we leave room to support protocols outside the Internet protocol family, and existing --ipv4 is grandfathered in by making it a synonym to --transfer-protocol-family=ipv4. Calling the variable "transfer.ipversion" and still allowing future protocols outside the Internet protocol family is rather awkward.

Calling "transfer.protocolFamily" would not have such a problem, though.

I agree that's a better name. I'm still on the fence about "transfer" versus "core".

-Peff

#35 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 3 lines · Jeff King

Adding a command-line option for "all" is a good idea, but will probably mean needing to add the "unset" sentinel value I mentioned in the other email.

Sorry, I do not quite follow. I thought that assigning the (misnamed --- see other mail) ALL to the "family" variable would be sufficient?

    enum transport_family {
            TRANSPORT_FAMILY_ALL = 0,
            TRANSPORT_FAMILY_IPV4,
            TRANSPORT_FAMILY_IPV6
    };
#36 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 10 lines · Junio C Hamano
Quoted 3 lines · Jeff King

Adding a command-line option for "all" is a good idea, but will probably mean needing to add the "unset" sentinel value I mentioned in the other email.

Sorry, I do not quite follow. I thought that assigning the (misnamed --- see other mail) ALL to the "family" variable would be sufficient?

Quoted 5 lines
    enum transport_family {
            TRANSPORT_FAMILY_ALL = 0,
            TRANSPORT_FAMILY_IPV4,
            TRANSPORT_FAMILY_IPV6
    };

Ah, I see. We want a way to tell "nobody has set it from the command line or the config" and "we were explicitly told to accept any" apart.

But wouldn't the usual "read config first and then override from the command line" handle that without "not yet set" value? I thought we by default accept any.

#37 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 14 lines · Junio C Hamano
Quoted 8 lines
Quoted 3 lines

Adding a command-line option for "all" is a good idea, but will probably mean needing to add the "unset" sentinel value I mentioned in the other email.

Sorry, I do not quite follow. I thought that assigning the (misnamed --- see other mail) ALL to the "family" variable would be sufficient?

Quoted 5 lines
    enum transport_family {
            TRANSPORT_FAMILY_ALL = 0,
            TRANSPORT_FAMILY_IPV4,
            TRANSPORT_FAMILY_IPV6
    };
Quoted 7 lines

Ah, I see. We want a way to tell "nobody has set it from the command line or the config" and "we were explicitly told to accept any" apart.

But wouldn't the usual "read config first and then override from the command line" handle that without "not yet set" value? I thought we by default accept any.

It would, if each individual program does it in that order. But that means every caller of the transport code needs to be updated to handle the config. That might not be that bad (after all, they have to take --ipv6 etc, options, and I guess they'd need a new "any" option).

My suggestion elsewhere was to have an "unset" value, and then resolve it at the time-of-use, something like:

transport.c +3 -0
index 43e24bf1e5..6414a847ae 100644
@@ -248,6 +248,9 @@ static int connect_setup(struct transport *transport, int for_push)
 	if (data->conn)
 		return 0;
 
+	if (transport->family == TRANSPORT_FAMILY_UNSET)
+		transport->family = transport_family_config;
+
 	switch (transport->family) {
 	case TRANSPORT_FAMILY_ALL: break;
 	case TRANSPORT_FAMILY_IPV4: flags |= CONNECT_IPV4; break;

but I am happy either way as long as the code does the right thing.

-Peff

#38 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 8 lines · Jeff King

My suggestion elsewhere was to have an "unset" value, and then resolve it at the time-of-use, something like:

transport.c +0 -0
index 43e24bf1e5..6414a847ae 100644

transport.c static int connect_setup(struct transport *transport, int for_push) · line 253

Quoted 6 lines
transport.c +3 -0
@@ -248,6 +248,9 @@ static int connect_setup(struct transport *transport, int for_push)
 	if (data->conn)
 		return 0;
 
+	if (transport->family == TRANSPORT_FAMILY_UNSET)
+		transport->family = transport_family_config;
+

Ah, OK, if we want to configure it the other way around, yes, we need "the command line didn't say any" value. The context of the "elsewhere" discussion was wnat I was missing (I'd happily blame vger for not delivering mails in order ;-).

#39 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

In reply to Junio C Hamano
Quoted 25 lines · Junio C Hamano

Also, we should follow the usual "the last one wins" for a configuration variable like this, which is *not* a multi-valued variable. So the config parsing would look more like this:

	if (!strcmp(k, "transfer.ipversion")) {
		if (!v)
			return config_error_nonbool("transfer.ipversion");
		if (!strcmp(v, "any"))
			family = 0;
		else if (!strcmp(v, "4") || !strcmp(v, "ipv4"))
			family = TRANSPORT_FAMILY_IPV4;
		else if (!strcmp(v, "6") || !strcmp(v, "ipv6"))
			family = TRANSPORT_FAMILY_IPV6;
		else
			return error("transfer.ipversion: unknown value '%s'", v);
	}

Would we regret to choose 'ipversion' as the variable name, by the way? On the command line side, --transfer-protocol-family=ipv4 makes it clear that we leave room to support protocols outside the Internet protocol family, and existing --ipv4 is grandfathered in by making it a synonym to --transfer-protocol-family=ipv4. Calling the variable "transfer.ipversion" and still allowing future protocols outside the Internet protocol family is rather awkward.

Quoted 2 lines

Calling "transfer.protocolFamily" would not have such a problem, though.

In case it wasn't clear, I consider the current TRANSPORT_FAMILY_ALL a misnomer. It's not like specifying "all" will make us use both ipv4 and ipv6 at the same time0---it just indicates our lack of preference, i.e. "any transport protocol family would do".

I mention this because this topic starts to expose that 'lack of preference' to the end user; I do not think we want to use "all" as the potential value for the command line option or the configuration variable.

Thanks.

#40 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Junio C Hamano, Wed, Sep 16, 2020 23:35:23 +0200:

Quoted 17 lines
Quoted 10 lines · Junio C Hamano

Would we regret to choose 'ipversion' as the variable name, by the way? On the command line side, --transfer-protocol-family=ipv4 makes it clear that we leave room to support protocols outside the Internet protocol family, and existing --ipv4 is grandfathered in by making it a synonym to --transfer-protocol-family=ipv4. Calling the variable "transfer.ipversion" and still allowing future protocols outside the Internet protocol family is rather awkward.

Calling "transfer.protocolFamily" would not have such a problem, though.

In case it wasn't clear, I consider the current TRANSPORT_FAMILY_ALL a misnomer. It's not like specifying "all" will make us use both ipv4 and ipv6 at the same time0---it just indicates our lack of preference, i.e. "any transport protocol family would do".

Quoted 4 lines

I mention this because this topic starts to expose that 'lack of preference' to the end user; I do not think we want to use "all" as the potential value for the command line option or the configuration variable.

If the configuration variable is allowed to be set to that "lack of preference" value, we kind of have a command line option for it:

git -c transfer.protocolFamily=any fetch ...

#41 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 4 lines · Alex Riesen

If the configuration variable is allowed to be set to that "lack of preference" value, we kind of have a command line option for it:

git -c transfer.protocolFamily=any fetch ...

Yes.

But we typically do not count that as being able to override from the command line. If "git fetch --ipv4" will defeat the configured "transfer.protoclFamily=ipv6", the users will expect there is some way to do the same and say they accept any protocol family to be used.

Thanks.

#42 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Junio C Hamano, Fri, Sep 18, 2020 00:31:58 +0200:

Quoted 7 lines
Quoted 4 lines · Alex Riesen

If the configuration variable is allowed to be set to that "lack of preference" value, we kind of have a command line option for it:

git -c transfer.protocolFamily=any fetch ...

Quoted 7 lines

Yes.

But we typically do not count that as being able to override from the command line. If "git fetch --ipv4" will defeat the configured "transfer.protoclFamily=ipv6", the users will expect there is some way to do the same and say they accept any protocol family to be used.

Makes sense. I even wondered myself why there is no way to override an --ipv4/--ipv6 in command-line back to "any" with an option added after it:

    $ cat my-fetch
    #!/bin/sh
    git fetch --ipv4 "$@"
    $ ./my-fetch --ipv46

But how about making the command-line and config option a list? (renaming it to ipversion, as elsewhere in the discussion)

    git fetch --ipversions=ipv6,ipv8

Given multiple times, the last option wins, as usual:

    $ cat my-fetch
    #!/bin/sh
    git fetch --ipversions=ipv4 "$@"
    $ ./my-fetch --ipversions=all

BTW, transport.c already converts transport->family to bit-flags in connect_setup.

#43 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 3 lines · Alex Riesen
    git fetch --ipversions=ipv6,ipv8

Given multiple times, the last option wins, as usual:

Just a clarification on "the last option wins".

You do not mean "I said v6 earlier but no, I want v8", with the above. What you mean is that

Quoted 6 lines
    $ cat my-fetch
    #!/bin/sh
    git fetch --ipversions=ipv4 "$@"
    $ ./my-fetch --ipversions=all

the argument given to 'my-fetch' overrides what is hardcoded in 'my-fetch', i.e. "I said v4, but I take it back; I want to accept any".

I find the above sensible.

Quoted 2 lines

BTW, transport.c already converts transport->family to bit-flags in connect_setup.

Yes, that is why I suggested the "list of acceptable choices" approach as a direction to go in the future, primarily to limit the scope of this current work. I do not mind it if you want to bite the whole piece right now, though.

By the way, I have a mild preference to call the option after the phrase "protocol-family", without "IP", so that we won't be limited to Internet protocols. IOW, --ipversions is a bad name for the new commnad line option in my mind.

As I said elsewhere, I also think TRANSPORT_FAMILY_ALL is a misnomer. When it is specified, we don't use all the available ones at the same time. What it says is that we accept use of any protocol families that are supported. It is OK to use ALL in the CPP macro as it is merely an internal implementation detail, but if we are going to expose it to end users as one of the choices, I'd prefer to use 'any', and not 'all', as the value for the new command line option.

Thanks.

#44 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Junio C Hamano, Fri, Sep 18, 2020 18:37:30 +0200:

Quoted 16 lines
Quoted 3 lines · Alex Riesen
    git fetch --ipversions=ipv6,ipv8

Given multiple times, the last option wins, as usual:

Just a clarification on "the last option wins".

You do not mean "I said v6 earlier but no, I want v8", with the above. What you mean is that

Quoted 5 lines
    $ cat my-fetch
    #!/bin/sh
    git fetch --ipversions=ipv4 "$@"
Quoted 5 lines
Quoted 1 line
    $ ./my-fetch --ipversions=all

the argument given to 'my-fetch' overrides what is hardcoded in 'my-fetch', i.e. "I said v4, but I take it back; I want to accept any".

Absolutely correct.

Quoted 1 line

I find the above sensible.

... And is precisely my intention.

Quoted 7 lines
Quoted 2 lines

BTW, transport.c already converts transport->family to bit-flags in connect_setup.

Yes, that is why I suggested the "list of acceptable choices" approach as a direction to go in the future, primarily to limit the scope of this current work. I do not mind it if you want to bite the whole piece right now, though.

I shall try, I think.

Quoted 4 lines

By the way, I have a mild preference to call the option after the phrase "protocol-family", without "IP", so that we won't be limited to Internet protocols. IOW, --ipversions is a bad name for the new commnad line option in my mind.

I have nothing against protocol-family, with or without "transport-". I just didn't want to ... over-generalize it prematurely: currently, the transport family is very fixed on IP on many levels.

Quoted 8 lines

As I said elsewhere, I also think TRANSPORT_FAMILY_ALL is a misnomer. When it is specified, we don't use all the available ones at the same time. What it says is that we accept use of any protocol families that are supported. It is OK to use ALL in the CPP macro as it is merely an internal implementation detail, but if we are going to expose it to end users as one of the choices, I'd prefer to use 'any', and not 'all', as the value for the new command line option.

Noted.

Regards,
Alex

#45 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

Quoted 8 lines · Alex Riesen
Quoted 4 lines

By the way, I have a mild preference to call the option after the phrase "protocol-family", without "IP", so that we won't be limited to Internet protocols. IOW, --ipversions is a bad name for the new commnad line option in my mind.

I have nothing against protocol-family, with or without "transport-". I just didn't want to ... over-generalize it prematurely: currently, the transport family is very fixed on IP on many levels.

I'll echo Junio's mild preference. I like it also because "--ipversion=6" feels too terse in the argument, but "--ipversion=ipv6" feels redundant. Saying "--protocol-family=ipv6" is more characters, but strikes me as more clear. (Or "transport-family" or whatever).

-Peff

#46 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

In reply to Junio C Hamano

Junio C Hamano, Wed, Sep 16, 2020 22:14:08 +0200:

Quoted 8 lines

How about introducing a new command line option

	--transfer-protocol-family=("any"|<protocol>)

where <protocol> is either "ipv4" or "ipv6" [*1*], and make existing "--ipv4" a synonym for "--transfer-protocol-family=ipv4" and "--ipv6" for "--transfer-protocol-family=ipv6"

Quoted 6 lines

With such an extended command line override, we can override configured

[transfer]
ipversion = 6

So the config option starts looking like "transfer.protocols" and multi-value? With the command-line option named "--transfer-protocol=", allowing multiple specification, and with "any" value taking precedence if specified anywhere?

#47 lore raw

Re: [PATCH] config: option transfer.ipversion to set transport protocol version for network fetches

In reply to Junio C Hamano

Junio C Hamano, Wed, Sep 16, 2020 22:14:08 +0200:

Quoted 11 lines
Quoted 5 lines · Alex Riesen

Affecting the transfers caused by git-fetch, the option allows to control network operations similar to --ipv4 and --ipv6 options.
...
Something like this?

Also, we should follow the usual "the last one wins" for a configuration variable like this, which is *not* a multi-valued variable. ...

...

Quoted 5 lines

[Footnote]

*1* But leave a room to extend it in the future to a comma-separated
list of them to allow something like "ipv6,ipv7,ipv8" (i.e. "not just 'any'---we want to say that 'ipv4' is not welcomed").

I think this footnote is the best description of this option. From what I gathered, it looks like really a list of protocols the networking code is allowed to try to reach the remote.

The above is orthogonal to how it given on the command-line: there it can be "unset". It is just not "unset" for the networking code, rather reset to the default list of protocols.

#48 lore raw
In reply to Alex Riesen
Quoted 15 lines · Alex Riesen
builtin/fetch.c +4 -0
index 82ac4be8a5..5e06c07106 100644
@@ -1531,6 +1531,10 @@ static void add_options_to_argv(struct argv_array *argv)
 		argv_array_push(argv, "-v");
 	else if (verbosity < 0)
 		argv_array_push(argv, "-q");
+	if (family == TRANSPORT_FAMILY_IPV4)
+		argv_array_push(argv, "--ipv4");
+	else if (family == TRANSPORT_FAMILY_IPV6)
+		argv_array_push(argv, "--ipv6");
 
 }
 
Quoted 1 line

Am I missing something obvious?

I think something obvious was missed back wne -4/-6 was added at c915f11e (connect & http: support -4 and -6 switches for remote operations, 2016-02-03) ;-).

The other candidate was 9c4a036b (Teach the --all option to 'git fetch', 2009-11-09) that introduced this helper to relay various options, but back then there weren't -4/-6 invented yet, so...

It is somewhat sad that we need to manually relay these down, but I do not offhand think of a way to automate this sensibly.

Thanks for noticing.