# meson build: datadir missing? 4 messages from 2025-03-03 to 2025-03-05. Participants: Steffen Nurpmeso, Patrick Steinhardt. Thread: https://gitlist.dev/t/63051 ## Steffen Nurpmeso, 2025-03-03 20:33 Subject: meson build: datadir missing? Message-ID: <20250303203320.lP10PHC5@steffen%sdaoden.eu> URL: https://gitlist.dev/e/20250303203320.lP10PHC5%40steffen%25sdaoden.eu ``` Hello again after a long time. I opened https://git.crux.nu/ports/opt/issues/18 at my linux distro ([ports/opt] git: creates incomplete repos), because hook etc directories were missing: warning: templates not found in /share/git-core/templates He then said Try setting [init] templatedir = /usr/share/git-core/templates in your gitconfig. which made me wonder since that is the default, so i looked and saw his change to switch to meson back in January. (What a pity, i liked the make system; only that it requires GNU make, but well..). I then said (it is a web ui, sorry): hi. thanks, i see now (after looking around). it is a fallout of switching to meson, that build system seems incomplete master:meson.build: '-DDEFAULT_GIT_TEMPLATE_DIR="' + get_option('datadir') / 'git-core/templates' + '"', but that "datadir" does not happen to be set at all it seems; is this known upstream? (Hamano does not like me either, so i will not do that..) Only wondering why this happens at all, and is not covered by its tests, really. Ok i overwrite the default with the real default in my config, thanks again. While here, since a couple of months maybe i get a problem on my spare notebook when i sync the filesystem (btrfs). To reproduce: #?0|kent:~# mount /media/btrfs-master/ #?0|kent:/media/btrfs-master# cd crux/kent/root #?0|kent:/media/btrfs-master/crux/kent/root# chroot . ^(we are now chroot(1)ed to the same what is /, actually.) #?0|kent:/# cd root/ #?0|kent:~# git co kent git: ../git-2.48.1/exec-cmd.c:50: system_prefix: Assertion `executable_dirname' failed. Aborted It will work if i do /usr/bin/git, effectively the same binary. However, this does not apply "normally": #?0|kent:~# cd / #?0|kent:/# chroot . #?0|kent:/# cd root #?0|kent:~# git co kent M hosts/iwd.network/.known_network.freq Switched to branch 'kent' #?0|kent:~# git co master M hosts/iwd.network/.known_network.freq Switched to branch 'master' Whatever the actual logic is, it seems a bit odd, and furthermore, and most of all, the above assertion is not a real help to solve the problem at hand, some kind of hint to a solution would be tremendous! Thanks for git, and ciao! --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt) ``` ## Patrick Steinhardt, 2025-03-04 07:08 Subject: Re: meson build: datadir missing? Message-ID: URL: https://gitlist.dev/e/Z8am0ZmkJ7QD13_g%40pks.im In-Reply-To: <20250303203320.lP10PHC5@steffen%sdaoden.eu> ``` On Mon, Mar 03, 2025 at 09:33:20PM +0100, Steffen Nurpmeso wrote: > Hello again after a long time. > > I opened > > https://git.crux.nu/ports/opt/issues/18 > > at my linux distro ([ports/opt] git: creates incomplete repos), > because hook etc directories were missing: > > warning: templates not found in /share/git-core/templates Okay. > hi. thanks, i see now (after looking around). it is a fallout of switching to meson, that build system seems incomplete > > master:meson.build: '-DDEFAULT_GIT_TEMPLATE_DIR="' + get_option('datadir') / 'git-core/templates' + '"', > > but that "datadir" does not happen to be set at all it seems; is this known upstream? (Hamano does not like me either, so i will not do that..) "datadir" is set, as it is a default option provided by Meson itself, and can be configured with `meson setup --datadir`. I've also double-checked the logic how we set the template directory path, but it does seem to be correct: our Makefile sets up a path relative to the prefix, and we do the exact same here. Double-checking on my machine also confirmed that this works as expected: $ meson setup /tmp/build --prefix=/tmp/prefix $ meson install -C /tmp/build $ touch /tmp/prefix/share/git-core/templates/foobar $ /tmp/prefix/bin/git git init /tmp/repo $ ls /tmp/repo/.git config description foobar HEAD hooks info objects refs reftable So I had a closer look at the recipe that Crux uses for building Git [1]. As it turns out, the issue isn't the prefix or datadir, but that Crux enables `-Druntime_prefix=true`: $ meson setup /tmp/build --prefix=/tmp/prefix -Druntime_prefix=true $ meson install -C /tmp/build $ /tmp/prefix/bin/git git init /tmp/repo warning: templates not found in /share/git-core/templates This is a bug that has already been fixed via bd262d07b65 (meson: fix exec path with enabled runtime prefix, 2025-02-26). I'd recommend to backport this fix or disable the runtime prefix. > Only wondering why this happens at all, and is not covered by its tests, really. Things like this are somewhat hard to test for as they require a proper installation. [1]: https://git.crux.nu/ports/opt/src/branch/3.7/git/Pkgfile Patrick ``` ## Steffen Nurpmeso, 2025-03-04 18:16 Subject: Re: meson build: datadir missing? Message-ID: <20250304181636.CEUTB9ed@steffen%sdaoden.eu> URL: https://gitlist.dev/e/20250304181636.CEUTB9ed%40steffen%25sdaoden.eu In-Reply-To: ``` Hello Patrick Steinhardt, Patrick Steinhardt wrote in : |On Mon, Mar 03, 2025 at 09:33:20PM +0100, Steffen Nurpmeso wrote: |> I opened |> |> https://git.crux.nu/ports/opt/issues/18 |> |> at my linux distro ([ports/opt] git: creates incomplete repos), |> because hook etc directories were missing: |> |> warning: templates not found in /share/git-core/templates | |Okay. | |> hi. thanks, i see now (after looking around). it is a fallout of \ |> switching to meson, that build system seems incomplete ... |> but that "datadir" does not happen to be set at all it seems; is \ ... |"datadir" is set, as it is a default option provided by Meson itself, |and can be configured with `meson setup --datadir`. I've also |double-checked the logic how we set the template directory path, but it |does seem to be correct: our Makefile sets up a path relative to the |prefix, and we do the exact same here. Double-checking on my machine |also confirmed that this works as expected: | | $ meson setup /tmp/build --prefix=/tmp/prefix | $ meson install -C /tmp/build | $ touch /tmp/prefix/share/git-core/templates/foobar | $ /tmp/prefix/bin/git git init /tmp/repo | $ ls /tmp/repo/.git | config description foobar HEAD hooks info objects refs reftable Thanks for all the effort! |So I had a closer look at the recipe that Crux uses for building Git |[1]. As it turns out, the issue isn't the prefix or datadir, but that |Crux enables `-Druntime_prefix=true`: | | $ meson setup /tmp/build --prefix=/tmp/prefix -Druntime_prefix=true | $ meson install -C /tmp/build | $ /tmp/prefix/bin/git git init /tmp/repo | warning: templates not found in /share/git-core/templates | |This is a bug that has already been fixed via bd262d07b65 (meson: fix |exec path with enabled runtime prefix, 2025-02-26). I'd recommend to |backport this fix or disable the runtime prefix. I will pass this information through, thank you. |> Only wondering why this happens at all, and is not covered by its \ |> tests, really. | |Things like this are somewhat hard to test for as they require a proper |installation. Sure. I just did not know about the meson stuff until this issue arose, lots of things seem to happen this year regarding it, there is a massive flux etc etc. I did not want to piss somewhere, sorry. |[1]: https://git.crux.nu/ports/opt/src/branch/3.7/git/Pkgfile | |Patrick --End of Thanks for git! Ciao, and greetings, --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt) ``` ## Patrick Steinhardt, 2025-03-05 07:19 Subject: Re: meson build: datadir missing? Message-ID: URL: https://gitlist.dev/e/Z8f6_3nJ9ld4kC6p%40pks.im In-Reply-To: <20250304181636.CEUTB9ed@steffen%sdaoden.eu> ``` On Tue, Mar 04, 2025 at 07:16:36PM +0100, Steffen Nurpmeso wrote: > Patrick Steinhardt wrote in > |> Only wondering why this happens at all, and is not covered by its \ > |> tests, really. > | > |Things like this are somewhat hard to test for as they require a proper > |installation. > > Sure. I just did not know about the meson stuff until this issue > arose, lots of things seem to happen this year regarding it, there > is a massive flux etc etc. I did not want to piss somewhere, > sorry. No need to be sorry, I wasn't offended by anything. Patrick ```