RFR: 8323672: Suppress unwanted autoconf added flags in CC and CXX [v7]

Julian Waters jwaters at openjdk.org
Tue Mar 12 15:10:21 UTC 2024


On Mon, 12 Feb 2024 14:09:49 GMT, Julian Waters <jwaters at openjdk.org> wrote:

>> I have verified that this works fine in the Oracle internal CI.
>> 
>> Erik's point still stands:
>>> I still think it would be prudent to verify this patch with all the currently accepted versions of autoconf (2.69, 2.70, 2.71, 2.72).
>> 
>> I think the easiest way to achieve this is to checkout the autoconf at these versions, and build it yourself. Iirc it was quite easy to build autoconf (anything else would be a shame and very bad PR for the project! :-o).
>> 
>> I'm hoping you are willing to do this, since I believe this is a superior solution and I'd like to see it in mainline. (Otherwise, let me know and I'll try to squeeze in doing it.)
>
>> I have verified that this works fine in the Oracle internal CI.
>> 
>> Erik's point still stands:
>> 
>> > I still think it would be prudent to verify this patch with all the currently accepted versions of autoconf (2.69, 2.70, 2.71, 2.72).
>> 
>> I think the easiest way to achieve this is to checkout the autoconf at these versions, and build it yourself. Iirc it was quite easy to build autoconf (anything else would be a shame and very bad PR for the project! :-o).
>> 
>> I'm hoping you are willing to do this, since I believe this is a superior solution and I'd like to see it in mainline. (Otherwise, let me know and I'll try to squeeze in doing it.)
> 
> It's actually even easier to do on Windows than it might appear, since MinGW's pacman system has caches that contain past package versions. I wouldn't have to build it at all in fact, all I'd have to do is downgrade my autoconf to past versions (I think I'm currently on 2.71). I know you're probably very busy right now, so I'll spare you the chore of having to do that by testing it on my end. That said, should I test autoconf on all platforms too? That aside, I'm still a little unhappy that there is no formal way to unregister an AC_DEFUN macro though :(
> 
> By the way, I've left you something in the mail. Hope you have some time to check it out!

> @TheShermanTanker I think this is essentially done. All that needs are the testing. Can you do it or do you want help with it?

I'm able to do so myself, I've just been busy with university so far. I've only tested 2.71 and 2.72 though, 2.69 and 2.70 might still need testing. I'm still unhappy that I can't properly undefine the macros, as a side tangent

-------------

PR Comment: https://git.openjdk.org/jdk/pull/17401#issuecomment-1991874882


More information about the build-dev mailing list