Skip to contentSkip to conversation

webgit highlightes mem adresses as git versions

PATCH In review Toralf Förster 7 participants 22 messages Jun 15, 2016

lore
Share & tools
lore
#1 lore raw
Quoted 4 lines · Toralf Förster

As seen here
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.27.y.git;a=commit;h=8ca2918f99b5861359de1805f27b08023c82abd2
the strings [<c043d0f3>] and firends shouldn't be recognized as git hashes, isn't it ?

Gitweb, not webgit. And gitweb considers ([0-9a-fA-F]{8,40}) i.e. from 8 to 40 hexadecimal characters to be (shortened) SHA-1. It simply cannot afford checking if such object exists when displaying commit message (for example in 'log' view).

Signature · 3 lines
Jakub Narebski
Poland
ShadeHawk on #git
#2 lore raw
Quoted 10 lines · At Monday 02 February 2009 23:54:20 Jakub Narebski wrote :
Quoted 4 lines · Toralf Förster

As seen here
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.27.y.git;a=com
mit;h=8ca2918f99b5861359de1805f27b08023c82abd2 the strings [<c043d0f3>] and firends shouldn't be recognized as git hashes, isn't it ?

Gitweb, not webgit. And gitweb considers ([0-9a-fA-F]{8,40}) i.e. from 8 to 40 hexadecimal characters to be (shortened) SHA-1. It simply cannot afford checking if such object exists when displaying commit message (for example in 'log' view).

Ah - ok, what's about expecting spaces around such SHA-1 keys ?

Signature · 4 lines
MfG/Sincerely

Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
#3 lore raw

Hi,

Quoted 7 lines · Toralf Förster
Quoted 6 lines · At Monday 02 February 2009 23:54:20 Jakub Narebski wrote :
Quoted 4 lines · Toralf Förster

As seen here
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.27.y.git;a=com
mit;h=8ca2918f99b5861359de1805f27b08023c82abd2 the strings [<c043d0f3>] and firends shouldn't be recognized as git hashes, isn't it ?

Quoted 6 lines
Quoted 4 lines

Gitweb, not webgit. And gitweb considers ([0-9a-fA-F]{8,40}) i.e. from 8 to 40 hexadecimal characters to be (shortened) SHA-1. It simply cannot afford checking if such object exists when displaying commit message (for example in 'log' view).

Ah - ok, what's about expecting spaces around such SHA-1 keys ?

Won't fly: there was a recommendation at some point that you should refer to commits in such a form:

	2819075(Merge branch 'maint-1.6.0' into maint)

However, gitweb being written in Perl, I think a lookbehind like (?<!0x), i.e. that a 0x in front of the hexadecimal characters means it is no SHA-1.

Even better would be using word boundaries, I guess, but all that fails when you have a hexdump in the commit message.

Ciao,
Dscho

#4 lore raw

PATCH 1/1 gitweb: Better regexp for SHA-1 committag match

Quoted 18 lines · Johannes Schindelin
Quoted 7 lines · Toralf Förster
Quoted 6 lines · At Monday 02 February 2009 23:54:20 Jakub Narebski wrote :
Quoted 3 lines · Toralf Förster

As seen here
http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.27.y.git;a=commit;h=8ca2918f99b5861359de1805f27b08023c82abd2 the strings [<c043d0f3>] and firends shouldn't be recognized as git hashes, isn't it ?

Quoted 6 lines
Quoted 4 lines

Gitweb, not webgit. And gitweb considers ([0-9a-fA-F]{8,40}) i.e. from 8 to 40 hexadecimal characters to be (shortened) SHA-1. It simply cannot afford checking if such object exists when displaying commit message (for example in 'log' view).

Ah - ok, what's about expecting spaces around such SHA-1 keys ?

Won't fly: there was a recommendation at some point that you should refer to commits in such a form:

Quoted 8 lines
	2819075(Merge branch 'maint-1.6.0' into maint)

However, gitweb being written in Perl, I think a lookbehind like (?<!0x), i.e. that a 0x in front of the hexadecimal characters means it is no SHA-1.

Even better would be using word boundaries, I guess, but all that fails when you have a hexdump in the commit message.

Here you have it: anchoring SHA-1 regexp to word boundary. It would help eliminate _some_ of false matches.

-- >8 --
Subject: [PATCH] gitweb: Better regexp for SHA-1 committag match

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

---
 gitweb/gitweb.perl |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
gitweb/gitweb.perl +1 -1
index f27dbb6..bec1af6 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
 	my $line = shift;
 
 	$line = esc_html($line, -nbsp=>1);
-	if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+	if ($line =~ m/\b([0-9a-fA-F]{8,40})\b/) {
 		my $hash_text = $1;
 		my $link =
 			$cgi->a({-href => href(action=>"object", hash=>$hash_text),
Signature · 1 line
1.6.1
#5 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Further suggestion: you could also turn the final \b into (\b|\@), so it skips stuff that might look like a message-id.
Here's an example :
http://perl5.git.perl.org/perl.git/commit/f57255841c18e91c7a719a2400645e39398f3947
(We get loads of this in the Perl repository)

Quoted 27 lines
---
 gitweb/gitweb.perl |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
gitweb/gitweb.perl +1 -2
index f27dbb6..bec1af6 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
       my $line = shift;
 
       $line = esc_html($line, -nbsp=>1);
-       if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+       if ($line =~ m/\b([0-9a-fA-F]{8,40})\b/) {
               my $hash_text = $1;
               my $link =
                       $cgi->a({-href => href(action=>"object", hash=>$hash_text),
--

1.6.1

--
To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 3 lines
"You don't mean odds and ends, you mean des curieux et des bouts",
corrected the manager.
-- Terry Pratchett, Hogfather
#6 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

Quoted 9 lines · Rafael Garcia-Suarez

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Further suggestion: you could also turn the final \b into (\b|\@),

You meant \b -> \b(?!\@), didn't you? Word boundary _not_ followed by '@', and not word boundary _OR_ '@' as you wrote...

Quoted 3 lines

so it skips stuff that might look like a message-id. Here's an example :
http://perl5.git.perl.org/perl.git/commit/f57255841c18e91c7a719a2400645e39398f3947

For those who do not want to open browser, it is:

Message-ID: <[email protected]>

Quoted 4 lines

(We get loads of this in the Perl repository)

Quoted 2 lines
gitweb/gitweb.perl +1 -1
-       if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+       if ($line =~ m/\b([0-9a-fA-F]{8,40})\b/) {
Signature · 2 lines
Jakub Narebski
Poland
#7 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 9 lines
Quoted 8 lines · Rafael Garcia-Suarez

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Quoted 4 lines
Quoted 1 line

Further suggestion: you could also turn the final \b into (\b|\@),

You meant \b -> \b(?!\@), didn't you? Word boundary _not_ followed by '@', and not word boundary _OR_ '@' as you wrote...

Ah right, shame on me.

#8 lore raw

PATCH v2 1/1 gitweb: Better regexp for SHA-1 committag match

Make SHA-1 regexp to be turned into hyperlink (SHA-1 committag) to match word boundary at beginning and end. This way we limit false matches, for example 0x74a5cd01 which is hex decimal (for example memory address) but not SHA-1.

Also make sure that it is not Message-ID, which fragment just looks like SHA-1 (e.g. "Message-ID: <[email protected]>"), by using zero-width negative look-ahead assertion to _not_ match '@' after.

---
v2: Added protection against matching Message-IDs fragments.

 gitweb/gitweb.perl |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
gitweb/gitweb.perl +1 -1
index f27dbb6..5dcc108 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
 	my $line = shift;
 
 	$line = esc_html($line, -nbsp=>1);
-	if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+	if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(!?\@)/) {
 		my $hash_text = $1;
 		my $link =
 			$cgi->a({-href => href(action=>"object", hash=>$hash_text),
Signature · 1 line
1.6.1
#9 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Hi,

gitweb/gitweb.perl sub format_log_line_html { · line 1367

Quoted 10 lines · Jakub Narebski
gitweb/gitweb.perl +1 -1
index f27dbb6..5dcc108 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
 	my $line = shift;
 
 	$line = esc_html($line, -nbsp=>1);
-	if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+	if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(!?\@)/) {

Looks good to me!

Thanks,
Dscho

#10 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 11 lines · Johannes Schindelin

Hi,

Quoted 7 lines · Jakub Narebski
gitweb/gitweb.perl +0 -0
index f27dbb6..5dcc108 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
      my $line = shift;
 
Quoted 5 lines

gitweb/gitweb.perl sub format_log_line_html { · line 1367

Quoted 3 lines
gitweb/gitweb.perl +1 -1
      $line = esc_html($line, -nbsp=>1);
-     if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+     if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(!?\@)/) {

Looks good to me!

I wonder if just matching lower-case a-f would be sufficient as well?

j.

#11 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 11 lines · Jay Soffian
Quoted 9 lines · Johannes Schindelin
Quoted 7 lines · Jakub Narebski
gitweb/gitweb.perl +0 -0
index f27dbb6..5dcc108 100755
@@ -1364,7 +1364,7 @@ sub format_log_line_html {
      my $line = shift;
 
Quoted 7 lines
Quoted 5 lines

gitweb/gitweb.perl sub format_log_line_html { · line 1367

Quoted 3 lines
gitweb/gitweb.perl +1 -1
      $line = esc_html($line, -nbsp=>1);
-     if ($line =~ m/([0-9a-fA-F]{8,40})/) {
+     if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(!?\@)/) {

Looks good to me!

I wonder if just matching lower-case a-f would be sufficient as well?

Well...

On one hand side git generates always lower-case a-f for SHA-1. On the other hand git _accepts_ upper-case A-F for SHA-1 of object.

Signature · 2 lines
Jakub Narebski
Poland
#12 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

In reply to Jakub Narebski

gitweb/gitweb.perl sub format_log_line_html { · line 1367

Quoted 1 line · Jakub Narebski
gitweb/gitweb.perl +1 -0
+	if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(!?\@)/) {

+ if ($line =~ m/\b([0-9a-fA-F]{8,40})\b(?!\@)/) {

Not that it matters, because adding such single-case exceptions is not a good idea, so it is v1 which would be (I hope) in git.

Signature · 2 lines
Jakub Narebski
Poland
#13 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

In reply to Jakub Narebski
Quoted 9 lines · Jakub Narebski

Make SHA-1 regexp to be turned into hyperlink (SHA-1 committag) to match word boundary at beginning and end. This way we limit false matches, for example 0x74a5cd01 which is hex decimal (for example memory address) but not SHA-1.

Also make sure that it is not Message-ID, which fragment just looks like SHA-1 (e.g. "Message-ID: <[email protected]>"), by using zero-width negative look-ahead assertion to _not_ match '@' after.

Your message I am responding to is:

    Message-ID: <[email protected]>

Does your description mean that "200902061149" would match, because the LAA will say "Ah, dot is not an at-sign"?

#14 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 12 lines · Junio C Hamano
Quoted 9 lines · Jakub Narebski

Make SHA-1 regexp to be turned into hyperlink (SHA-1 committag) to match word boundary at beginning and end. This way we limit false matches, for example 0x74a5cd01 which is hex decimal (for example memory address) but not SHA-1.

Also make sure that it is not Message-ID, which fragment just looks like SHA-1 (e.g. "Message-ID: <[email protected]>"), by using zero-width negative look-ahead assertion to _not_ match '@' after.

Quoted 6 lines

Your message I am responding to is:

    Message-ID: <[email protected]>

Does your description mean that "200902061149" would match, because the LAA will say "Ah, dot is not an at-sign"?

It would unfortunately falsely match... but we cannot eliminate this case (well, at least not checking if hexnumber is followed by dot), because of totally legitimate

... at commit 8457bb9e.

So even with that we would have still false matches...

Signature · 2 lines
Jakub Narebski
Poland
#15 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 7 lines · Jakub Narebski

It would unfortunately falsely match... but we cannot eliminate this case (well, at least not checking if hexnumber is followed by dot), because of totally legitimate

   ... at commit 8457bb9e.

So even with that we would have still false matches...

Yeah, so what's the value in v2 over v1? It is still wrong but it is less wrong than it used to be? I think the word-boundary one made a good sense. I do not see the @lookahead adding much value at all.

#16 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 6 lines · Junio C Hamano
Quoted 4 lines · Jakub Narebski

It would unfortunately falsely match... but we cannot eliminate this case (well, at least not checking if hexnumber is followed by dot), because of totally legitimate

Quoted 7 lines
Quoted 3 lines
   ... at commit 8457bb9e.

So even with that we would have still false matches...

Yeah, so what's the value in v2 over v1? It is still wrong but it is less wrong than it used to be? I think the word-boundary one made a good sense. I do not see the @lookahead adding much value at all.

Right. So v2 is less useful that I thought it to be; and adding further "exceptions" doesn't seem like a good idea. The 'msgid' committag when/if it gets implemented would help there...

So please take v1, as it is sane improvement and generic enough.

Signature · 2 lines
Jakub Narebski
Poland
#17 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

2009/2/7 Jakub Narebski <[email protected]>:

Quoted 15 lines
Quoted 6 lines · Junio C Hamano
Quoted 4 lines · Jakub Narebski

It would unfortunately falsely match... but we cannot eliminate this case (well, at least not checking if hexnumber is followed by dot), because of totally legitimate

Quoted 7 lines
Quoted 3 lines
   ... at commit 8457bb9e.

So even with that we would have still false matches...

Yeah, so what's the value in v2 over v1? It is still wrong but it is less wrong than it used to be? I think the word-boundary one made a good sense. I do not see the @lookahead adding much value at all.

Quoted 5 lines

Right. So v2 is less useful that I thought it to be; and adding further "exceptions" doesn't seem like a good idea. The 'msgid' committag when/if it gets implemented would help there...

So please take v1, as it is sane improvement and generic enough.

If you make it configurable then everybody can be happy right?

Yves

Signature · 1 line
perl -Mre=debug -e "/just|another|perl|hacker/"
#18 lore raw

Re: [PATCHv2] gitweb: Better regexp for SHA-1 committag match

Quoted 16 lines · demerphq

2009/2/7 Jakub Narebski <[email protected]>:

Quoted 15 lines
Quoted 13 lines · Junio C Hamano
Quoted 7 lines · Jakub Narebski

It would unfortunately falsely match... but we cannot eliminate this case (well, at least not checking if hexnumber is followed by dot), because of totally legitimate

   ... at commit 8457bb9e.

So even with that we would have still false matches...

Yeah, so what's the value in v2 over v1? It is still wrong but it is less wrong than it used to be? I think the word-boundary one made a good sense. I do not see the @lookahead adding much value at all.

Quoted 7 lines
Quoted 5 lines

Right. So v2 is less useful that I thought it to be; and adding further "exceptions" doesn't seem like a good idea. The 'msgid' committag when/if it gets implemented would help there...

So please take v1, as it is sane improvement and generic enough.

If you make it configurable then everybody can be happy right?

That are the long term plans, to implement generic 'committags' support (which would include current SHA-1 and signoff committags).

BTW. it is the 'configurable' part that makes it difficult... ;-)

Signature · 2 lines
Jakub Narebski
Poland
#19 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

In reply to Jakub Narebski

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 9 lines
Quoted 8 lines · Rafael Garcia-Suarez

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Quoted 4 lines
Quoted 1 line

Further suggestion: you could also turn the final \b into (\b|\@),

You meant \b -> \b(?!\@), didn't you? Word boundary _not_ followed by '@', and not word boundary _OR_ '@' as you wrote...

Since \b(?!\@) is effectively two zero width negative assertions in a row you could simplify by saying:

(?![^\w\@])

and that way you can easily add the '.' case as well.

Yves

Signature · 1 line
perl -Mre=debug -e "/just|another|perl|hacker/"
#20 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

Quoted 12 lines · demerphq

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 11 lines
Quoted 9 lines · Rafael Garcia-Suarez

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Further suggestion: you could also turn the final \b into (\b|\@),

Quoted 7 lines
Quoted 2 lines

You meant \b -> \b(?!\@), didn't you? Word boundary _not_ followed by '@', and not word boundary _OR_ '@' as you wrote...

Since \b(?!\@) is effectively two zero width negative assertions in a row you could simplify by saying:

(?![^\w\@])

I don't know if "sth\b" is effectively "sth(!?[^\w])"... perhaps it is.

Quoted 2 lines

and that way you can easily add the '.' case as well.

We cannot add '.' case, because it there can be legitimate SHA-1 match ending sentence, e.g.

     ... at commit 8457bb9e.
Signature · 2 lines
Jakub Narebski
Poland
#21 lore raw

Re: [PATCH] gitweb: Better regexp for SHA-1 committag match

2009/2/7 Jakub Narebski <[email protected]>:

Quoted 16 lines · demerphq
Quoted 15 lines

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 13 lines
Quoted 9 lines · Rafael Garcia-Suarez

2009/2/6 Jakub Narebski <[email protected]>:

Quoted 5 lines

Make SHA-1 regexp to be turned into hyperlink (the SHA-1 committag) to match word boundary at the beginning and the end. This way we reduce number of false matches, for example we now don't match 0x74a5cd01 which is hex decimal (for example memory address), but is not SHA-1.

Further suggestion: you could also turn the final \b into (\b|\@),

You meant \b -> \b(?!\@), didn't you? Word boundary _not_ followed by '@', and not word boundary _OR_ '@' as you wrote...

Quoted 6 lines
Quoted 4 lines

Since \b(?!\@) is effectively two zero width negative assertions in a row you could simplify by saying:

(?![^\w\@])

I don't know if "sth\b" is effectively "sth(!?[^\w])"... perhaps it is.

Sorry, my bad, that is double negation, I meant (?![\w\@])

On of the ways you can express \b is as:
(?:(?<=\w)(?!\w)|(?<=\W)(?!\W)|\A)

But the point here is you are looking for the end of a hex sequence, so you can just use the "end of string" bit of the alternation which
is: (?!\w).

Quoted 7 lines
Quoted 2 lines

and that way you can easily add the '.' case as well.

We cannot add '.' case, because it there can be legitimate SHA-1 match ending sentence, e.g.

    ... at commit 8457bb9e.

/(?<!\w)([a-fA-F0-9]+)(?!(?:\.\w|[\w@]))/

:-)

Yves

Signature · 1 line
perl -Mre=debug -e "/just|another|perl|hacker/"