Re: Git ChangeLog policy for GCC Testsuite inquiry

[Segher Boessenkool <segher@kernel.crashing.org>]

On Thu, Feb 06, 2020 at 03:56:40PM +0100, Jakub Jelinek wrote:
> On Wed, Feb 05, 2020 at 06:43:54PM -0700, Jeff Law wrote:
> > And FWIW, we're talking about the ChangeLog *file* here.  If folks
> > continued writing the same log messages and put them into git, I
> > personally think that's sufficient to transition away from having a
> > ChangeLog file in the source tree.
> 
> A precondition for such change would be a script that would verify
> that each commit being pushed into the repository has the ChangeLog snippets
> available.  And not just that.  With ChangeLog files, one can fix up the
> entries later on if somebody makes a mistake; it happens from time to time
> to everybody, some people make the same mistakes again and again.

Yes.  Mistakes *will* happen, and for *everyone*.

> If the ChangeLog entry is in the commit message, then in git (unlike svn) it
> is not possible to fix it later once you push it into the public repository.
> 
> So, we'd need a script that checks for the common mistakes, furthermore
> verifies that e.g. all files in the commit that need to have a ChangeLog
> entry have one etc. and reports back any issues + rejects the push.

We also need a way to fix changelog entries for the errors that do seep
through (and that are bad enough that they do need fixing).  It doesn't
have to be easy or convenient, but we need *some* way to do it.

> We would need to agree how do we express stuff going into different former
> ChangeLog files, whether we require gcc/cp/ etc. prefixes before the lines,
> or say require empty line for different former ChangeLog files and let the
> extraction script figure it out from the changed files in the commit, etc.

The latter seems error-prone, but maybe I worry too much :-)

> We could have a rule that say New test. entries for newly added files
> in */testsuite/* would be optional and let the generating script add those
> (but say require entries if existing tests have been changed).

How much work does that safe?

> Plus the extraction script, and test properly both scripts for a while
> before making it live.

Yes, absolutely.  Maybe it could post it to gcc-cvs@ (as it will in the
future) (but not actually commit the (generated) changelog parts yet)?


Segher