automatic CC generation for patch submission
Hello,
I'd like to present a suggestion for automatic generation of carbon copy fields in the E-Mails of posted patches.
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking at MAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
To illustrate: If a patch affects a file under drivers/net/e1000, the CC script will look at these files
drivers/net/e1000/.maintainers
drivers/net/.maintainers
drivers/.maintainers
.maintainers
... to gather up the mailing list addresses or an individual maintainer inbox address.
A posssible format for this file could be a newline-separated list of:
[filename wildcard]:e-mail
For example, drivers/scsi/.maintainers would contain:
libiscsi.*:[email protected]
scsi_*.c:[email protected]
etc...
Or, instead (or in addition) of having a '.maintainers' file each directory we can modify source files by adding parsable '/* MAINTAINER: name@domain */' comments.
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
* Date: Sat, 30 Jun 2007 05:34:51 +0300
Quoted 11 lines
Hello,
I'd like to present a suggestion for automatic generation of carbon copy fields in the E-Mails of posted patches.
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking at
MAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
LKML archive near you, search phrase "BTS", subjects that have something about "the quality of the kernel". Good luck.
____
Description in other words. Forget about scripts in source tree, tracking and management system (as already described) is needed.
--
-o--=O`C
#oo'L O
<___=E M
Quoted 7 lines · Oleg Verych
* Date: Sat, 30 Jun 2007 05:34:51 +0300
Quoted 6 lines
Hello,
I'd like to present a suggestion for automatic generation of carbon copy fields in the E-Mails of posted patches.
Quoted 8 lines
Quoted 5 lines
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking at
MAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.LKML archive near you, search phrase "BTS", subjects that have something about "the quality of the kernel". Good luck.
Right, however many patches don't map to bug reports and don't need the heavy use of BTS. This suggestion is mainly for the improvement of peer review concerning code changes submitted by people who are not the maintainers.
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
Quoted 14 lines · Dan Aloni
Quoted 13 lines · Oleg Verych
* Date: Sat, 30 Jun 2007 05:34:51 +0300
Quoted 11 lines
Hello,
I'd like to present a suggestion for automatic generation of carbon copy fields in the E-Mails of posted patches.
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking at
MAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
Quoted 7 lines
Quoted 2 lines
LKML archive near you, search phrase "BTS", subjects that have something about "the quality of the kernel". Good luck.
Right, however many patches don't map to bug reports and don't need the heavy use of BTS. This suggestion is mainly for the improvement of peer review concerning code changes submitted by people who are not the maintainers.
Unfortunately you didn't listen, OK, just two messages:
<http://permalink.gmane.org/gmane.linux.kernel/544898>
<http://permalink.gmane.org/gmane.linux.kernel/545112>
As you will see, nobody cares about comprehensive patches/tests/bugs/testers/developers *tracking* system.
And don't limit yourself to fast conclusions. Thanks.
____
Quoted 7 lines · Oleg Verych
Quoted 4 lines
Right, however many patches don't map to bug reports and don't need the heavy use of BTS. This suggestion is mainly for the improvement of peer review concerning code changes submitted by people who are not the maintainers.
Unfortunately you didn't listen, OK, just two messages:
Quoted 7 lines
<http://permalink.gmane.org/gmane.linux.kernel/544898>
<http://permalink.gmane.org/gmane.linux.kernel/545112>As you will see, nobody cares about comprehensive patches/tests/bugs/testers/developers *tracking* system.
And don't limit yourself to fast conclusions. Thanks.
I am not proposing a comprehensive tracking system. I think you are taking my intentions quite off-course.
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
On Sat, Jun 30, 2007 at 01:26:01PM +0300, Dan Aloni wrote:
[]
Quoted 7 lines
Quoted 4 lines
As you will see, nobody cares about comprehensive patches/tests/bugs/testers/developers *tracking* system.
And don't limit yourself to fast conclusions. Thanks.
I am not proposing a comprehensive tracking system. I think you are taking my intentions quite off-course.
Don't you think that organized (i.e. Cc *management*: tracking system sends copies to _interested people_, as it was recorded earlier, or changed due to bouncing or off-line people; by-patch tracking of changes, tests, bugs, rc-bugs) way of dealing with patches is bad, but yet another fancy SPAM generator script and mess in ___kernel sources___ is OK just for publishing one's patches-ches-es?
Having seen how bitkeeper improved *management* things in 2.5 - early2.6 period, i have no problem to blame propositions as yours.
Heck, going to be another useless post...
---
-o--=O`C
#oo'L O
<___=E M
Quoted 28 lines · Dan Aloni
Hello,
I'd like to present a suggestion for automatic generation of carbon copy fields in the E-Mails of posted patches.
Basically, instead of manually figuring out who to add to CC
when sending a patch to LKML by looking atMAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs
for a patch.To illustrate: If a patch affects a file under
drivers/net/e1000, the CC script will look at these files
drivers/net/e1000/.maintainers
drivers/net/.maintainers
drivers/.maintainers
.maintainers... to gather up the mailing list addresses or an individual maintainer inbox address.
A posssible format for this file could be a newline-separated list of:
[filename wildcard]:e-mail
For example,
drivers/scsi/.maintainerswould contain:libiscsi.*:[email protected]
scsi_*.c:[email protected]
Quoted 7 lines
etc...
Or, instead (or in addition) of having a '.maintainers' file each directory we can modify source files by adding parsable '/* MAINTAINER: name@domain */' comments.
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
an easier way to implement this is to add an extra field in the MAINTAINERS file, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Auke
(horribly whitespace-mutilated copy+paste from thunderbird below)
---
MAINTAINERS +4 -0
@@ -69,6 +69,7 @@ L: Mailing list that is relevant to this area W: Web-page with status/info T: SCM tree type and location. Type is one of: git, hg, quilt. S: Status, one of the following: +F: Directory tree or Files belonging to this subsystem Supported: Someone is actually paid to look after this. Maintained: Someone actually looks after it.
@@ -1880,6 +1881,7 @@ M: [email protected] L: [email protected] W: http://sourceforge.net/projects/e1000/ S: Supported +F: drivers/net/e100.c INTEL PRO/1000 GIGABIT ETHERNET SUPPORT P: Jeb Cramer
@@ -1895,6 +1897,7 @@ M: [email protected] L: [email protected] W: http://sourceforge.net/projects/e1000/ S: Supported +F: drivers/net/e1000/ INTEL PRO/10GbE SUPPORT P: Jeff Kirsher
@@ -1910,6 +1913,7 @@ M: [email protected] L: [email protected] W: http://sourceforge.net/projects/e1000/ S: Supported +F: drivers/net/ixgb/ INTEL PRO/WIRELESS 2100 NETWORK CONNECTION SUPPORT P: Yi Zhu
Quoted 4 lines · Auke
[...]
an easier way to implement this is to add an extra field in theMAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Hmm, associating MAINTAINERS and source files. I remember surfing LKML archives one day and coming across this huge and ugly flamefest when some guy called Eric S. Raymond (hehe :-) suggested something like this and got burned.
Cheers,
S
Quoted 5 lines · Satyam Sharma
Quoted 4 lines · Auke
[...]
an easier way to implement this is to add an extra field in theMAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Your idea is quite different though, of course.
Quoted 4 lines
Hmm, associating
MAINTAINERSand source files. I remember surfing LKML archives one day and coming across this huge and ugly flamefest when some guy called Eric S. Raymond (hehe :-) suggested something like this and got burned.
<ugh, reads again and feels stupid>
And before _I_ get burned for being impolite / provoking something, let me clarify that the comment above was purely in spirit of humour (bad effort, of course).
Quoted 10 lines · Auke
Quoted 6 lines
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
an easier way to implement this is to add an extra field in the
MAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
We already have that information in git. Parse the git changelogs of the affected files, find out who works on them.
Not that it'll help much, given the amnount of stuff which gets mysteriously ignored even when the correct people are cc'ed...
(For extra giggles we could parse emailed oops and bug reports and add the appropriate cc's there too. Harder.)
Quoted 7 lines · Andrew Morton
Quoted 5 lines · Auke
Quoted 5 lines
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Any comments?
an easier way to implement this is to add an extra field in the
MAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.We already have that information in git. Parse the git changelogs of the affected files, find out who works on them.
I think it's quite complex to make a reliable inference of maintainership information from git. Given a set of historical modifiers of a file, would you take the most common commiter(s), or the most common _recent_ commiter(s), or what? It's a bit fuzzy.
Moreover, it is slow in comparison and assumes the availability of local .git db, which wouldn't be the case for some porition of patch submitters.
Quoted 2 lines
Not that it'll help much, given the amnount of stuff which gets mysteriously ignored even when the correct people are cc'ed...
Hopefully it gets ignored if it is quite low in priority. In that case the CC is a NOP but might still be good for archiving purposes.
Quoted 2 lines
(For extra giggles we could parse emailed oops and bug reports and add the appropriate cc's there too. Harder.)
BTS was discussed to death already, let's not delve into that...
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
Quoted 14 lines · On Sat, 30 Jun 2007 13:47:52 +0300 Dan Aloni <[email protected]> wrote:
Quoted 13 lines · Andrew Morton
Quoted 10 lines · Auke
Quoted 6 lines
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
an easier way to implement this is to add an extra field in the
MAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Quoted 5 lines
Quoted 2 lines
We already have that information in git. Parse the git changelogs of the affected files, find out who works on them.
I think it's quite complex to make a reliable inference of maintainership information from git.
Clever people will work it out.
Quoted 3 lines
Given a set of historical modifiers of a file,
would you take the most common commiter(s), or the most common _recent_ commiter(s), or what? It's a bit fuzzy.
All the above? Multiply frequency by recency, pick the top five?
Doesn't matter much: the cost of picking too many is high.
I shudder at the thought of manually maintaining anything like this.
Quoted 3 lines
Moreover, it is slow in comparison and assumes the availability of local .git db, which wouldn't be the case for some porition of patch submitters.
a) precalculate the tables once per week
b) the whole thing wouldn't succeed if it requires software at patch-submitter's site. It'd need to run at vger.
Quoted 16 lines · Andrew Morton
Quoted 14 lines · On Sat, 30 Jun 2007 13:47:52 +0300 Dan Aloni <[email protected]> wrote:
Quoted 13 lines · Andrew Morton
Quoted 10 lines · Auke
Quoted 6 lines
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
an easier way to implement this is to add an extra field in the
MAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
We already have that information in git. Parse the git changelogs of the affected files, find out who works on them.
I think it's quite complex to make a reliable inference of maintainership information from git.
Clever people will work it out.
I don't see how git could tell you to Cc net patches to [email protected].
Quoted 9 lines
Quoted 3 lines
Given a set of historical modifiers of a file,
would you take the most common commiter(s), or the most common _recent_ commiter(s), or what? It's a bit fuzzy.All the above? Multiply frequency by recency, pick the top five?
Doesn't matter much: the cost of picking too many is high.
I shudder at the thought of manually maintaining anything like this.
Auke suggestion of an additional field in MAINTAINERS makes sense - and it leaves it to the maintainers to maintain it similar to the other information there.
Independent of whatever software Dan wants to write it would also solve problems like the one that MAINTAINERS currently isn't really good in telling e.g. who is maintaining drivers/net/wireless/.
Quoted 8 lines
Quoted 3 lines
Moreover, it is slow in comparison and assumes the availability of local .git db, which wouldn't be the case for some porition of patch submitters.
a) precalculate the tables once per week
b) the whole thing wouldn't succeed if it requires software at patch-submitter's site. It'd need to run at vger.
It might depend on what use cases you think of and whom you expect to use it.
For me, a tool that simply tells me whom to Cc for a patch would be a great tool.
What you suggest would be something completely different.
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 11 lines · Andrew Morton
[..]
Quoted 3 lines
Given a set of historical modifiers of a file,
would you take the most common commiter(s), or the most common _recent_ commiter(s), or what? It's a bit fuzzy.All the above? Multiply frequency by recency, pick the top five?
Doesn't matter much: the cost of picking too many is high.
I shudder at the thought of manually maintaining anything like this.
Why? It shouldn't take much more effort than the effort currently being taken to maintain the MAINTAINERS file as it is. This field can even be considered optional - i.e. let the maintainers themselves who usually send patches to MAINTAINERS just to update their mailing address add this field _only if they want to_.
Heck, this way it would even help us to see which maintainers are more attentive :)
Quoted 9 lines
Quoted 3 lines
Moreover, it is slow in comparison and assumes the availability of local .git db, which wouldn't be the case for some porition of patch submitters.
a) precalculate the tables once per week
b) the whole thing wouldn't succeed if it requires software at patch-submitter's site. It'd need to run at vger.
And:
c) The script will need to query vger for each file in the patch via some protocol..
Do you think it's worth the effort?
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
Quoted 7 lines · Andrew Morton
Quoted 4 lines · Auke
Quoted 3 lines
Some extensions to the popular E-Mail clients might be needed here. Also, a bot reading LKML would automatically send links about posted patches to the other mailing lists whenever someone forgets to add a CC.
Any comments?
an easier way to implement this is to add an extra field in the
MAINTAINERSfile, something like below. All the contact info would stay the same, closely where applicable and it would allow you to also specify specific files as well.
Quoted 8 lines
We already have that information in git. Parse the git changelogs of the affected files, find out who works on them.
Not that it'll help much, given the amnount of stuff which gets mysteriously ignored even when the correct people are cc'ed...
(For extra giggles we could parse emailed oops and bug reports and add the appropriate cc's there too. Harder.)
well, I'm more thinking of the advertisement value of having some sort of pointer in the MAINTAINERS file to the exact location of a certain subsystem. Quite often I find myself searching somewhere in /drivers/ only to find out that the stuff I need is in asm/arch or elsewhere. E.g. alsa is a driver but lives in /sound/. Having a voluntary field (usable for other purposes) in the MAINTAINERS field seems awfully appropriate in that place and very low impact.
git does have that info, except for the fact that quite often others touch the same subsystem much more than the maintainer (pci, net, etc)...
Auke
Quoted 15 lines · Dan Aloni
...
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking atMAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
...
To illustrate: If a patch affects a file underdrivers/net/e1000, the CC script will look at these files
drivers/net/e1000/.maintainers
drivers/net/.maintainers
drivers/.maintainers
.maintainers
Quoted 4 lines
... to gather up the mailing list addresses or an individual maintainer inbox address.
...
Any comments?
As Auke said, maintaining the information in MAINTAINERS would be better.
And another important use case that shouldn't require much extra work would be to do the same for bug reports.
Generally, you should keep in mind that it must fit into the workflow of the people who should use it. E.g. I could imagine that it might make more sense if you write a small tool that takes a patch or a path and outputs email addresses instead of a huge tool that tries to solve too many problems at once and doesn't fit into the workflow of most people.
Quoted 1 line
Dan Aloni
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon SeedQuoted 16 lines · Adrian Bunk
Quoted 15 lines · Dan Aloni
...
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking atMAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
...
To illustrate: If a patch affects a file underdrivers/net/e1000, the CC script will look at these files
drivers/net/e1000/.maintainers
drivers/net/.maintainers
drivers/.maintainers
.maintainers
Quoted 7 lines
Quoted 4 lines
... to gather up the mailing list addresses or an individual maintainer inbox address.
...
Any comments?As Auke said, maintaining the information in
MAINTAINERSwould be better.
Yes, I tend to agree.
Quoted 2 lines
And another important use case that shouldn't require much extra work would be to do the same for bug reports.
Yes, however for bug reports it isn't always possible to tell which subsystem or files would get fixed once the bug is understood. Perhaps it would be good once the bug report reaches the stage where patches are made.
Quoted 5 lines
Generally, you should keep in mind that it must fit into the workflow of the people who should use it. E.g. I could imagine that it might make more sense if you write a small tool that takes a patch or a path and outputs email addresses instead of a huge tool that tries to solve too many problems at once and doesn't fit into the workflow of most people.
I absolutely agree - in fact I was thinking of this as a nice addition to scripts/checkpatch.pl. For example:
Your patch has no obvious style problems and is ready for submission. When submitting this particular patch, you should CC to:
[email protected], [email protected]
However my Perl hacking capabilities (or incapabilities thereof) wouldn't do good to that script I suppose.
Signature · 3 lines
Dan Aloni XIV LTD, http://www.xivstorage.com da-x (at) monatomic.org, dan (at) xiv.co.il
Quoted 27 lines · Dan Aloni
Quoted 16 lines · Adrian Bunk
Quoted 15 lines · Dan Aloni
...
Basically, instead of manually figuring out who to add to CC when sending a patch to LKML by looking atMAINTAINERS, a script can look at '.maintainers' files spread across the source tree and automatically generate a proper list of CCs for a patch.
...
To illustrate: If a patch affects a file underdrivers/net/e1000, the CC script will look at these files
drivers/net/e1000/.maintainers
drivers/net/.maintainers
drivers/.maintainers
.maintainersQuoted 7 lines
Quoted 4 lines
... to gather up the mailing list addresses or an individual maintainer inbox address.
...
Any comments?As Auke said, maintaining the information in
MAINTAINERSwould be better.Yes, I tend to agree.
Quoted 7 lines
Quoted 2 lines
And another important use case that shouldn't require much extra work would be to do the same for bug reports.
Yes, however for bug reports it isn't always possible to tell which subsystem or files would get fixed once the bug is understood. Perhaps it would be good once the bug report reaches the stage where patches are made.
Don't aim for perfection.
If your program also accepts a path or file as input (instead of extracting it from a patch) you'll get it with nearly zero extra effort.
Quoted 6 lines
Quoted 5 lines
Generally, you should keep in mind that it must fit into the workflow of the people who should use it. E.g. I could imagine that it might make more sense if you write a small tool that takes a patch or a path and outputs email addresses instead of a huge tool that tries to solve too many problems at once and doesn't fit into the workflow of most people.
Quoted 7 lines
I absolutely agree - in fact I was thinking of this as a nice addition to
scripts/checkpatch.pl. For example:Your patch has no obvious style problems and is ready for submission. When submitting this particular patch, you should CC to:
checkpatch.pl checks a patch, not an email.
And e.g. in my workflow there isn't any technical correlation between some "Cc:" field in a patch and the recipients of the email containing it since I'm not using any fancy scripts for maintaining and sending patches.
Quoted 2 lines
However my Perl hacking capabilities (or incapabilities thereof) wouldn't do good to that script I suppose.
A tool that simply would output some email addresses to Cc when given a patch is something I would most likely use since it would fit in my workflow and replace manual looks at MAINTAINERS.
Something big is less likely to be adapted unless it nearly perfectly suits the needs of the people intended to use it.
But it's your program, so it's your choice what you implement and in what language you implement it.
Quoted 1 line
Dan Aloni
cu
Adrian
Signature · 4 lines
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed