This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Move stuff to simplify-rtx.c, 6/13 - PLUS



If you're confident that the first five parts are safe, you can go
ahead and check them into mainline as independent commits.  But
I'm now cautious about proceeding further without better testing.
Moving this amount of code around in the RTL optimizers is tricky
business.

Yes, I know. :-( That's why I bootstrapped on two architectures and not just one.

I appreciate your work in reviewing these patches, and in catching the problem with a variable that is not declared. However, I did say that I bootstrapped the patch set all together, and that the split was only for review -- you can be confident that GCC contributors tend not to lie and not to post patches with syntax errors in them. :-) Diego once told me that splitting a patch so that only the full patch set compiles is acceptable, and I already did it with macho cleanup patches such as the one to move RTL passes under the pass manager.

If you prefer them to be committed independently, so be it, but it will require 13 bootstrap/regtest cycles, which means 1.5*13 machine hours without Java and with checking disabled. :-( Actually even more, because there's no guarantee that all bootstraps succeed at the first time; that I can, every 1.5 hours, interrupt what I'm doing to commit a patch and start a new test cycle; that nobody breaks bootstrap in the meanwhile.

What I'm inclined to ask for, is this: on one hand, you approve the patch set to commit it in one go; on the other hand, I'll wait to commit this patch set until the dust settles from Joern's cfgcleanup patch, and if it causes bootstrapping problems, everybody can send an e-mail to me with the .i file and revert my patch (without waiting 48 hours as the rules would say).

The force_to_mode patch, instead, was bootstrapped separately (one architecture only) and I can commit it.

Paolo

ps: CCing was not to stimulate an immediate review, of course, but just to let you know that the patch was split :-)


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]