On Fri, 12 Jan 2024 16:15:34 GMT, Julian Waters <jwat...@openjdk.org> wrote:
>> [JDK-8323008](https://bugs.openjdk.org/browse/JDK-8323008) reports unwanted >> autoconf flags being added to the command line, and solves the issue by >> filtering out the added flags by force. This is not optimal however, as >> doing so leaves the autoconf message that the flags were added still >> displaying during the configure process, which is confusing. Instead, >> setting a couple of fields to disable the autoconf internals responsible for >> the unwanted flag is a cleaner solution > > Julian Waters has updated the pull request incrementally with one additional > commit since the last revision: > > make/autoconf/toolchain.m4 > > Co-authored-by: Magnus Ihse Bursie <m...@icus.se> The current use of AC_PROG_CC does not fit very well with out needs. It does a lot of testing that is not needed for us (and at least on Windows, this considerably slows down configure). So it would make sense to replace it with something that only does what we need. I have thought of this before, even without the context of this bug. However, I have not pushed in this direction yet, partly because of lack of time, but also partly since I was afraid (well-grounded or not) that this was doing a lot of "autoconf black magic" that is needed elsewhere. But if you can take the time to analyze and understand what the vanilla AC_PROG_CC macro actually does, and extract just the parts that are relevant for us, and argue convincingly that the parts we're leaving out is not going to affect us, then I think it would be good. The more I've worked with autoconf, the less happy I've become with all its magic stuff, unchangable premises and bad fit for our use case. ------------- PR Comment: https://git.openjdk.org/jdk/pull/17401#issuecomment-1891942358
