[WIP] meson: update to 1.11.1.

[sjozs]

Testing the changes

• I tested the changes in this PR: YES
  I built a randomly selected subset of about 30 x86_64-glibc packages with checks (I got one failure on libgweather, which has a broken network file dependency in the tests, otherwise everything worked). Then I tried cross building using my new meson package to x86_64-musl, i686 and aarch64 . This also worked. (Some were only a single build, since I figured that if mesa compiles, everything compiles and building it four times took about three hours alone.)

I have removed the rust-proc-macro patch since it's fixed in upstream now, and built a few build_style=meson packages that had cargo or rust somewhere in the template with the new changes to make sure they worked (with xbps-src -Q install). gst-plugins-rs1 does not compile, but it's a Rust error for implicit raw pointer dereference and it doesn't build with the old meson either, which is probably worth another issue (though it also has updates available, so I'm not sure what exactly to write in it, also it's not an orphaned package). Every other tested meson+Rust package builds successfully.

Local build testing

• I built this PR locally for my native architecture, x86_64-glibc (pure Python with no compiled parts, the only compiled code is in the tests that won't run anyway, so I did not try building for other archs).

(I found #58641 after I started testing my changes and it seemed better to simply open a new PR instead of trying to work on that since there was a version bump since that one. I also used the new meson to build and run a somewhat nontrivial very experimental autotools port of the very old game warmux to test.)

[sjozs]

Testing update: I tried rebuilding every meson-related (meson/mesonpy in dependencies or build_style/build_helper=meson) package, I have 41 (mostly check) failures from 640 so far (only x86_64, sorry). The total is somewhere in the neighborhood of 900, and I'd like to look at the failed stuff more, although the ones I've checked manually so far break regardless of the meson version, which looks promising. I'll update this comment when I see the results.

Edit: I got the results:

libucontext - musl-only
sagemath - broken doctests
python3-pywt - broken with old meson, fixed in HEAD
tinysparql - tests broken regardless of meson
iputils - can't open socket on localhost (no cap_net_raw+p)
newsflash - bad checksum
gst-plugins-rs1 - completely borked
switchboard - bad checksum
librist - tests are flaky, independent of meson
virt-manager - CLI test failure, breaks with old meson
open-isns - no idea but tests break with old meson as well
patchmatrix - distfiles moved
umockdev - out of date, Python deprecation warning
localsearch - missing checkdepends maybe?
gxml - breaks with both versions
switchboard-plug-bluetooth - bad checksum
appstream-glib - broken consistently
vali - 'NDEBUG must be disabled for tests' (if I disable it, both meson versions work)
gamescope - https://github.com/ValveSoftware/gamescope/issues/2110 (fixed in new releases)
gdk-pixbuf - test timeout
sagemath - doctests are generally broken, but it builds and works fine
pango - same broken tests
libpisp - no Intel archs
PackageKit - 'FATAL-WARNING: you need to compile with -Dlocal_checkout=true for ninja test support'
accountsservice - tests don't build (tries wrap download) - strange
switchboard-plug-network - bad checksum
libdazzle - works on second try
pcsc-ccid - missing distfiles
citations - metainfo warnings
rubberband - does not play nice with boost-headers (meson version does not matter)
fontmanager - https://github.com/FontManager/font-manager/issues/467 (does not compile with new Vala)
gnome-calculator - https://gitlab.gnome.org/GNOME/gnome-calculator/-/work_items/531 (mathematically equivalent results, tests break, now fixed in upstream, no release yet)
libwacom - can't create evdev devices in xbps-src, works with both meson versions outside
libvarlink - tests break the same way
libgweather - https://gitlab.gnome.org/GNOME/libgweather/-/work_items/348 (metar network dependency changed)
pacman - tests broken in the same way with both mesons
io.elementary.files - very old version broken with new gcc+vala (https://github.com/elementary/files/issues/2508)
ddgtk - appstream metadata does not validate (in either meson version)
clevis - missing checkdepends, fixed
gnome-authenticator - warnings in stock metainfo.xml (meson-independent)
corecollector - missing checkdepends, fixed
xorg-server - fails with old meson too (GLX missing error in build environment, not sure why or how)
d-feet - code style error
p11-kit - tests broken, related to suid, might be fixed in new release (https://github.com/p11-glue/p11-kit/issues/316#issuecomment-678302296)
nix - this is the only actually broken thing I found, will be fixed for real in 0.12, 
kiwix-lib - upstream bug, fixed in https://github.com/kiwix/libkiwix/issues/1291 (has new release)
switchboard-plug-printers - bad checksum

There is a regression in this version that breaks nix compilation, which got fixed in 1.11.2 no, it's mostly just MacOS and Windows fixes, which got bumped an hour ago and not yet tagged.

[sjozs]

I know it's not the best form to mix up two package updates in one PR, sorry, but this is kind of connected. For some reason I couldn't fix the nix version in the repos, but at least the new one compiled fine with the patch from the Fedora bug. I'm somewhat ambivalent on this update now, since this patch is at best a hack (and I can't fix meson instead since the patch would probably be larger than whatever is in the Firefox package). At least now everything works the same as before. The next person to update meson should probably fix nix (I can do it, if no one else steps up, at least I now have all my test scripts). If it's a problem, I can separate them somehow, not sure which is worse. I guess I could just ignore the problems with nix since I'm not using it, and the next person who cares about it can fix it, but at least here's something that seems to work. I'd say this is as ready as it can be. At least every other package builds fine (on x86_64, but the sample is probably not too small on other archs as well.)

And apparently, the nix manual only compiles when cross-compiling, sorry. I'll try to fix it. (And now it doesn't even do that, I probably messed up something and left a good executable in the masterdir then.)