Re: [gem5-dev] commit message thrown away because of bad tag
I've created a Jira ticket for this: https://gem5.atlassian.net/browse/GEM5-321 I'll get to work on it. -- Dr. Bobby R. Bruce Room 2235, Kemper Hall, UC Davis Davis, CA, 95616 web: https://www.bobbybruce.net On Mon, Feb 10, 2020 at 5:58 AM Jason Lowe-Power wrote: > Hi all, > > I think we should solve this problem instead ignoring it. It's good to have > this check as a huge number of reviews are simply "please update the commit > message", which is also a bad experience. > > @Bobby: Could you try to figure out a more user-friendly way for the commit > message to be rejected. IMO, this should work like a rejected rebase. If > you just do `git commit` again, then the original commit message should > still be there. > > Thanks! > Jason > > On Sun, Feb 9, 2020 at 11:47 PM Gabe Black wrote: > > > It is thrown away if you then attempt to create another commit, assuming > > the first one went worked, overwriting COMMIT_EDITMSG. Then when I wrote > > the message again it failed again, and then I ran git commit trying to > > figure out what happened, and the COMMIT_EDITMSG was thrown away again. > > There is also no way that I know of to bring that back into the editor to > > fix it. I have disabled the check locally and plan to leave it that way. > > > > Gabe > > > > On Sun, Feb 9, 2020 at 7:56 PM Bobby Bruce wrote: > > > > > Gabe, > > > > > > Commit messages are not thrown away but are instead stored in > > > "./.git/COMMIT_EDITMSG" > > > if rejected. Perhaps this should be explicitly stated when a commit > > message > > > is rejected. > > > > > > As you may already be aware, this is a result of Daniel's (relatively) > > new > > > git commit message checks: > > > > > > > > > https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py > > > . It seems the valid arm tag is "arch-arm". You could update your > commits > > > with this tag or add submit a change adding "arm" as a valid tag in > this > > > file (to lead to further discussion). > > > > > > I personally prefer there being a valid set of tags to stop the > constant > > > invention of new tags to describe commits which may, in practise, be > > > duplicates of old tags (though we can expand them as needed). I'm > > therefore > > > still in support of these checks. It also helps us understand who the > > > maintainers for a particular commits are by cross-referencing with > > > MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). > > > > > > Kind regards, > > > Bobby > > > -- > > > Dr. Bobby R. Bruce > > > Room 2235, > > > Kemper Hall, UC Davis > > > Davis, > > > CA, 95616 > > > > > > web: https://www.bobbybruce.net > > > > > > > > > On Sun, Feb 9, 2020 at 7:27 PM Gabe Black > wrote: > > > > > > > I think until this changes, --no-verify is going to become a > permanent > > > > fixture of my commit command line. > > > > > > > > Gabe > > > > > > > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black > > wrote: > > > > > > > > > Hi folks. I just spent 10 minutes writing up a commit message, only > > for > > > > it > > > > > to be thrown away because git didn't like the tag I had which I've > > used > > > > > many times (arm). This is a really bad experience and IMHO not how > > this > > > > > should be done. > > > > > > > > > > I believe at the time this change was introduced I pointed out how > > this > > > > > sort of check was a bit heavy handed, and I especially feel that > way > > > > after > > > > > it threw away my work. > > > > > > > > > > Gabe > > > > > > > > > ___ > > > > gem5-dev mailing list > > > > gem5-dev@gem5.org > > > > http://m5sim.org/mailman/listinfo/gem5-dev > > > ___ > > > gem5-dev mailing list > > > gem5-dev@gem5.org > > > http://m5sim.org/mailman/listinfo/gem5-dev > > ___ > > gem5-dev mailing list > > gem5-dev@gem5.org > > http://m5sim.org/mailman/listinfo/gem5-dev > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
Hi all, I think we should solve this problem instead ignoring it. It's good to have this check as a huge number of reviews are simply "please update the commit message", which is also a bad experience. @Bobby: Could you try to figure out a more user-friendly way for the commit message to be rejected. IMO, this should work like a rejected rebase. If you just do `git commit` again, then the original commit message should still be there. Thanks! Jason On Sun, Feb 9, 2020 at 11:47 PM Gabe Black wrote: > It is thrown away if you then attempt to create another commit, assuming > the first one went worked, overwriting COMMIT_EDITMSG. Then when I wrote > the message again it failed again, and then I ran git commit trying to > figure out what happened, and the COMMIT_EDITMSG was thrown away again. > There is also no way that I know of to bring that back into the editor to > fix it. I have disabled the check locally and plan to leave it that way. > > Gabe > > On Sun, Feb 9, 2020 at 7:56 PM Bobby Bruce wrote: > > > Gabe, > > > > Commit messages are not thrown away but are instead stored in > > "./.git/COMMIT_EDITMSG" > > if rejected. Perhaps this should be explicitly stated when a commit > message > > is rejected. > > > > As you may already be aware, this is a result of Daniel's (relatively) > new > > git commit message checks: > > > > > https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py > > . It seems the valid arm tag is "arch-arm". You could update your commits > > with this tag or add submit a change adding "arm" as a valid tag in this > > file (to lead to further discussion). > > > > I personally prefer there being a valid set of tags to stop the constant > > invention of new tags to describe commits which may, in practise, be > > duplicates of old tags (though we can expand them as needed). I'm > therefore > > still in support of these checks. It also helps us understand who the > > maintainers for a particular commits are by cross-referencing with > > MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). > > > > Kind regards, > > Bobby > > -- > > Dr. Bobby R. Bruce > > Room 2235, > > Kemper Hall, UC Davis > > Davis, > > CA, 95616 > > > > web: https://www.bobbybruce.net > > > > > > On Sun, Feb 9, 2020 at 7:27 PM Gabe Black wrote: > > > > > I think until this changes, --no-verify is going to become a permanent > > > fixture of my commit command line. > > > > > > Gabe > > > > > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black > wrote: > > > > > > > Hi folks. I just spent 10 minutes writing up a commit message, only > for > > > it > > > > to be thrown away because git didn't like the tag I had which I've > used > > > > many times (arm). This is a really bad experience and IMHO not how > this > > > > should be done. > > > > > > > > I believe at the time this change was introduced I pointed out how > this > > > > sort of check was a bit heavy handed, and I especially feel that way > > > after > > > > it threw away my work. > > > > > > > > Gabe > > > > > > > ___ > > > gem5-dev mailing list > > > gem5-dev@gem5.org > > > http://m5sim.org/mailman/listinfo/gem5-dev > > ___ > > gem5-dev mailing list > > gem5-dev@gem5.org > > http://m5sim.org/mailman/listinfo/gem5-dev > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
Gabe, I just saw your other e-mail (could you please CC me if you respond to this? My e-mail provider blocks all your e-mails, not even letting reach the spam box...), and although I did not understand the proper sequence of commands done, and I have never had an issue when creating commits after this commit, I will point you to a resource describing how to create a helpful alias in case that pops up frequently to you: https://stackoverflow.com/questions/52975103/how-can-i-recover-the-commit-message-when-the-git-commit-msg-hook-fails Try to not give up on these checks, since they standardize the codebase, which makes it easier for users to selectively apply modifications to their local gem5. Also, further improvements that depend on the validity of the commit could be made so that the maintainers are automatically assigned as reviewers given the informed tags, or even make these tags be added automatically. Regards, Daniel Em segunda-feira, 10 de fevereiro de 2020 08:57:22 GMT+1, Daniel Carvalho escreveu: Hi Gabe, Sorry for the inconvenience, but as Bobby pointed out, the commit message is not discarded, and its backup's location is explicitly stated on a failure. If it is preferred it could also be printed for ease of access, but when I tried this locally it clogged the output, making it harder to identify why the failure happened. Extract of the error message: " Invalid commit header The commit has been canceled, but a copy of it can be found in .git/COMMIT_EDITMSG [...]" Regards, Daniel Em segunda-feira, 10 de fevereiro de 2020 04:56:07 GMT+1, Bobby Bruce escreveu: Gabe, Commit messages are not thrown away but are instead stored in "./.git/COMMIT_EDITMSG" if rejected. Perhaps this should be explicitly stated when a commit message is rejected. As you may already be aware, this is a result of Daniel's (relatively) new git commit message checks: https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py . It seems the valid arm tag is "arch-arm". You could update your commits with this tag or add submit a change adding "arm" as a valid tag in this file (to lead to further discussion). I personally prefer there being a valid set of tags to stop the constant invention of new tags to describe commits which may, in practise, be duplicates of old tags (though we can expand them as needed). I'm therefore still in support of these checks. It also helps us understand who the maintainers for a particular commits are by cross-referencing with MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). Kind regards, Bobby -- Dr. Bobby R. Bruce Room 2235, Kemper Hall, UC Davis Davis, CA, 95616 web: https://www.bobbybruce.net On Sun, Feb 9, 2020 at 7:27 PM Gabe Black wrote: > I think until this changes, --no-verify is going to become a permanent > fixture of my commit command line. > > Gabe > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black wrote: > > > Hi folks. I just spent 10 minutes writing up a commit message, only for > it > > to be thrown away because git didn't like the tag I had which I've used > > many times (arm). This is a really bad experience and IMHO not how this > > should be done. > > > > I believe at the time this change was introduced I pointed out how this > > sort of check was a bit heavy handed, and I especially feel that way > after > > it threw away my work. > > > > Gabe > > > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
Hi Gabe, Sorry for the inconvenience, but as Bobby pointed out, the commit message is not discarded, and its backup's location is explicitly stated on a failure. If it is preferred it could also be printed for ease of access, but when I tried this locally it clogged the output, making it harder to identify why the failure happened. Extract of the error message: " Invalid commit header The commit has been canceled, but a copy of it can be found in .git/COMMIT_EDITMSG [...]" Regards, Daniel Em segunda-feira, 10 de fevereiro de 2020 04:56:07 GMT+1, Bobby Bruce escreveu: Gabe, Commit messages are not thrown away but are instead stored in "./.git/COMMIT_EDITMSG" if rejected. Perhaps this should be explicitly stated when a commit message is rejected. As you may already be aware, this is a result of Daniel's (relatively) new git commit message checks: https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py . It seems the valid arm tag is "arch-arm". You could update your commits with this tag or add submit a change adding "arm" as a valid tag in this file (to lead to further discussion). I personally prefer there being a valid set of tags to stop the constant invention of new tags to describe commits which may, in practise, be duplicates of old tags (though we can expand them as needed). I'm therefore still in support of these checks. It also helps us understand who the maintainers for a particular commits are by cross-referencing with MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). Kind regards, Bobby -- Dr. Bobby R. Bruce Room 2235, Kemper Hall, UC Davis Davis, CA, 95616 web: https://www.bobbybruce.net On Sun, Feb 9, 2020 at 7:27 PM Gabe Black wrote: > I think until this changes, --no-verify is going to become a permanent > fixture of my commit command line. > > Gabe > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black wrote: > > > Hi folks. I just spent 10 minutes writing up a commit message, only for > it > > to be thrown away because git didn't like the tag I had which I've used > > many times (arm). This is a really bad experience and IMHO not how this > > should be done. > > > > I believe at the time this change was introduced I pointed out how this > > sort of check was a bit heavy handed, and I especially feel that way > after > > it threw away my work. > > > > Gabe > > > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
It is thrown away if you then attempt to create another commit, assuming the first one went worked, overwriting COMMIT_EDITMSG. Then when I wrote the message again it failed again, and then I ran git commit trying to figure out what happened, and the COMMIT_EDITMSG was thrown away again. There is also no way that I know of to bring that back into the editor to fix it. I have disabled the check locally and plan to leave it that way. Gabe On Sun, Feb 9, 2020 at 7:56 PM Bobby Bruce wrote: > Gabe, > > Commit messages are not thrown away but are instead stored in > "./.git/COMMIT_EDITMSG" > if rejected. Perhaps this should be explicitly stated when a commit message > is rejected. > > As you may already be aware, this is a result of Daniel's (relatively) new > git commit message checks: > > https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py > . It seems the valid arm tag is "arch-arm". You could update your commits > with this tag or add submit a change adding "arm" as a valid tag in this > file (to lead to further discussion). > > I personally prefer there being a valid set of tags to stop the constant > invention of new tags to describe commits which may, in practise, be > duplicates of old tags (though we can expand them as needed). I'm therefore > still in support of these checks. It also helps us understand who the > maintainers for a particular commits are by cross-referencing with > MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). > > Kind regards, > Bobby > -- > Dr. Bobby R. Bruce > Room 2235, > Kemper Hall, UC Davis > Davis, > CA, 95616 > > web: https://www.bobbybruce.net > > > On Sun, Feb 9, 2020 at 7:27 PM Gabe Black wrote: > > > I think until this changes, --no-verify is going to become a permanent > > fixture of my commit command line. > > > > Gabe > > > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black wrote: > > > > > Hi folks. I just spent 10 minutes writing up a commit message, only for > > it > > > to be thrown away because git didn't like the tag I had which I've used > > > many times (arm). This is a really bad experience and IMHO not how this > > > should be done. > > > > > > I believe at the time this change was introduced I pointed out how this > > > sort of check was a bit heavy handed, and I especially feel that way > > after > > > it threw away my work. > > > > > > Gabe > > > > > ___ > > gem5-dev mailing list > > gem5-dev@gem5.org > > http://m5sim.org/mailman/listinfo/gem5-dev > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
Gabe, Commit messages are not thrown away but are instead stored in "./.git/COMMIT_EDITMSG" if rejected. Perhaps this should be explicitly stated when a commit message is rejected. As you may already be aware, this is a result of Daniel's (relatively) new git commit message checks: https://gem5.googlesource.com/public/gem5/+/refs/heads/master/util/git-commit-msg.py . It seems the valid arm tag is "arch-arm". You could update your commits with this tag or add submit a change adding "arm" as a valid tag in this file (to lead to further discussion). I personally prefer there being a valid set of tags to stop the constant invention of new tags to describe commits which may, in practise, be duplicates of old tags (though we can expand them as needed). I'm therefore still in support of these checks. It also helps us understand who the maintainers for a particular commits are by cross-referencing with MAINTAINERS.md (fyi: there is no maintainer for the "arm" tag). Kind regards, Bobby -- Dr. Bobby R. Bruce Room 2235, Kemper Hall, UC Davis Davis, CA, 95616 web: https://www.bobbybruce.net On Sun, Feb 9, 2020 at 7:27 PM Gabe Black wrote: > I think until this changes, --no-verify is going to become a permanent > fixture of my commit command line. > > Gabe > > On Sun, Feb 9, 2020 at 7:20 PM Gabe Black wrote: > > > Hi folks. I just spent 10 minutes writing up a commit message, only for > it > > to be thrown away because git didn't like the tag I had which I've used > > many times (arm). This is a really bad experience and IMHO not how this > > should be done. > > > > I believe at the time this change was introduced I pointed out how this > > sort of check was a bit heavy handed, and I especially feel that way > after > > it threw away my work. > > > > Gabe > > > ___ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
Re: [gem5-dev] commit message thrown away because of bad tag
I think until this changes, --no-verify is going to become a permanent fixture of my commit command line. Gabe On Sun, Feb 9, 2020 at 7:20 PM Gabe Black wrote: > Hi folks. I just spent 10 minutes writing up a commit message, only for it > to be thrown away because git didn't like the tag I had which I've used > many times (arm). This is a really bad experience and IMHO not how this > should be done. > > I believe at the time this change was introduced I pointed out how this > sort of check was a bit heavy handed, and I especially feel that way after > it threw away my work. > > Gabe > ___ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev