Mark Kettenis <mark.kette...@xs4all.nl> writes: >> From: Eric Anholt <e...@anholt.net> >> Date: Fri, 24 Mar 2017 13:17:45 -0700 >> >> Having bitten off a bit more than I can chew in 3 days with the X >> Server (hw/xfree86/sdksyms.c is the worst), I decided to take a quick >> pass at converting a project that's my own fault. > > Seems I missed some discussion somewhere... > > While I understand your frustrations with autoconf, moving to build > infrastructure that relies on tools like Python and Ninja would have > serious consequences for the way we integrate X into OpenBSD. We > build the entire base OS, which includes X, with tools that are part > of the base OS. That would become pretty much impossible if various X > projects go this way.
While I understand your frustrations with new build infrastructure requiring changes in your OS to integrate new versions of software, it has serious consequences for projects (as I mentioned, CI has been intractable with autoconf) to be stuck with bad tools because of your OS's choices. > This is the first I've heard about Meson. That in itself doesn't fill > me with confidence. But the Meson documentation doesn't really > mentioned how you'd do configure-style checks like checking whether a > particular header file is present or whether a library defines a > certain function. I suspect the answer is that people using Meson > don't do such tests and don't feel the need to do such tests because > they just check whether you're running on Windows, MacOS X or Linux > and make assumptions that things are present based on just that. Your > new meson build system doesn't check whether the err.h file exists or > whether the C compiler supports the -Wshadow option for example. It's > not really surprising that Meson is much faster if you cut out all > those checks. Unfortunately this approach tends to yield rather > unportable software. You don't need to check for -Wshadow because it's not on by default (it's a bad warning). If you want conditional cflags, they're easy. The err.h check I apparently just mis-grepped when seeing if HAVE_ERR_H was used. v2 has the trivial fix (timing is indistinguishable without rigging up compare-perf). > You can probably cut quite a bit of time from the autoconf configure > script run time by removing silly checks like checking for standard C > headers and such. As I said in the cover letter, rendercheck is the best possible case I know of for autotools, and autotools sucks at it.
signature.asc
Description: PGP signature
_______________________________________________ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: https://lists.x.org/mailman/listinfo/xorg-devel