add meson #1868

neheb · 2021-08-30T02:52:19Z

Signed-off-by: Rosen Penev rosenp@gmail.com

Gonna just post this here. I've tried to make it as similar as the CMakeLists.txt file.

Benefit is smaller size when built with minsize. It also works with coverity's build tool. The CMake one does not for some reason.

Some bugs are fixed compared to the cmake one: libatomic does not get linked unconditionally, the exiv2 auto_ptr hack is "fixed". The actual define needs to match the libc++, not the compiler.

FILE_OFFSET_BITS are also set to 64. Not configurable. ccache is default unless CC/CXX is overridden.

Various hacks still need to be fixed. Notably the git and compile info stuff.

whyman · 2021-08-30T06:47:46Z

I would rather have fixes for the CMake.

neheb · 2021-09-02T04:40:52Z

Added fixes from Debian 10. Turns out, even clang 7 can compile Gerbera.

KarlStraussberger · 2021-09-02T11:18:43Z

Please, add a section to doc/compile to explain usage.

neheb · 2021-09-06T06:30:31Z

Added doc section. Also updated to install files in /usr/share as well.

What's left with the meson build is:

testing support
proper support of compile info. It's currently hardcoded.

The big change is that it's actually usable now. GNOME Builder runs it just fine. Profiling also works.

edit: forgot to mention. I've only done testing on various Linux distros. No testing has been done on macOS or any BSD. MinGW was tested briefly but doesn't compile for unrelated reasons (bad std::fs usage, POSIX APIs, etc...)

eli-schwartz

Hi, nice to see you considering meson. Some suggestions for potential optimizations:

eli-schwartz · 2021-09-13T04:08:45Z

meson.build

+
+iconv_dep = cpp.find_library('iconv', required : false)
+if not iconv_dep.found()
+  if cpp.has_function('iconv')


In meson 0.60, you will be able to do dependency('iconv') as I've added a custom lookup in mesonbuild/meson#9205 which will first check if iconv_open is available in libc without any dependencies, and then look for cpp.find_library('iconv').

Unfortunately that does require a version so new it is not even released right now. But it will eventually let you simplify these 8 lines down to 1.

I'll keep this in mind.

meson.build

neheb · 2021-09-13T06:07:10Z

I think I addressed most stuff.

Some stuff I kept as booleans. Specifically the stuff that needs two or more dependencies.

neheb · 2021-09-13T06:11:52Z

Also note this project uses cxxopts as a single header library. Would it make sense to use https://github.com/mesonbuild/wrapdb/blob/master/subprojects/cxxopts.wrap ?

eli-schwartz · 2021-09-13T06:18:57Z

It would, but on the other hand I don't know if you can actually remove the cmake vendored copy just yet...

Either way, meson would be able to take advantage of encapsulation, accountability, and clean upgradeability of the .wrap file interface.

neheb · 2021-09-13T06:21:43Z

https://github.com/gerbera/gerbera/tree/master/scripts contains a bunch of scripts to install newer versions of packages. I assume these "wraps" build and statically link these libraries? If so, that would be a great alternative.

eli-schwartz · 2021-09-13T06:26:52Z

Indeed they do, with -Default_library=static or per-wrap -D<wrapname>:default_library=static meson enforces that those wraps are linked statically.

And they don't even need root in order to do it :p they are just downloaded and checksummed and private copies built. Automatically as part of searching for dependencies, if system copies are not found or if --wrap-mode=forcefallback forces wraps to be preferred.
(Those scripts do not do checksumming, BTW.)

neheb · 2021-09-13T06:33:14Z

I'll test later on Debian 10. fmtlib and spdlog are the issues there.

edit: first problem I see is the massive amount of -Wpsabi warnings on 32-bit ARM :).

neheb · 2021-09-13T07:52:11Z

Hmm doesn't seem to work with meson 0.49.1 (version that comes with Debian 10)

eli-schwartz · 2021-09-13T13:28:29Z

For versions of meson earlier than 0.55, automatic wrap fallback via the [provide] section was not yet implemented and that section of the .wrap file will be ignored.

If you need to support versions older than that, the dependency() function takes a kwarg fallback: [ 'subproject_name', 'foo_dep' ] which will fallback to configuring the subproject and looking for the variable foo_dep as the right-hand side of the mapping in the newer style [provide] section.

meson.build

neheb · 2021-09-14T03:53:25Z

Added compile_info. Not totally happy with how it turned out...

eli-schwartz · 2021-09-14T04:24:15Z

meson.build

+  uuid_dep,
+]
+
+compile_info = [ '-DCOMPILE_INFO="' +


This looks painful and complicated indeed. But this seems to be an array with one element, and later inserted as '-DCOMPILE_INFO='.join(compile_info)

I think maybe you meant to not include all those + ' ' +. Just have:

'WITH_MAGIC=' + magic_dep.found().to_string().to_upper(), 'WITH_MYSQL=' + mysql_dep.found().to_string().to_upper(), 'WITH_CURL=' + curl_dep.found().to_string().to_upper(),

etc.

maybe...

foreach info: [ ['MAGIC', magic_dep], ['MYSQL', mysql_dep], ... ] found = info[1].found().to_string().to_upper() compile_info += 'WITH_@0@=@1@'.format(info[0], found) endforeach

I'm very curious why the current cmake build includes COMPILE_INFO about whether or not the tests were enabled. :/ Likewise about whether the systemd unit was installed...

Given most of these are nearly identical to existing libgerbera_args += '-DHAVE_*' it feels like this should be handled in the .cpp source code using preprocessor macros to extend a string. But clearly that is not the current state of the cmake...

I tried the first version. I got output like -DCOMPILE_INFO="-DXXX=TRUE-DYY=FALSE-DZZZ=TRUE

Ah no

-DCOMPILE_INFO=\"WITH_MAGIC=TRUE-DCOMPILE_INFO=WITH_MYSQL=TRUE-DCOMPILE_INFO=WITH_CURL=TRUE-DCOMPILE_INFO=WITH_INOTIFY=TRUE-DCOMPILE_INFO=WITH_JS=TRUE-DCOMPILE_INFO=WITH_TAGLIB=TRUE-DCOMPILE_INFO=WITH_AVCODEC=TRUE-DCOMPILE_INFO=WITH_FFMPEGTHUMBNAILER=TRUE-DCOMPILE_INFO=WITH_EXIF=TRUE-DCOMPILE_I NFO=WITH_EXIV2=TRUE-DCOMPILE_INFO=WITH_SYSTEMD=TRUE-DCOMPILE_INFO=WITH_LASTFM=TRUE-DCOMPILE_INFO=WITH_DEBUG=TRUE-DCOMPILE_INFO=WITH_TESTS=FALSE\"'

well, '-DCOMPILE_INFO='.join(compile_info) will get you:

just compile_info, if there is only one element

otherwise: compile_info[0] + '-DCOMPILE_INFO=' + compile_info[1] + '-DCOMPILE_INFO=' + compile_info[2] .........

But you have a second copy of -DCOMPILE_INFO= at the beginning of your compile_info string building. So you might as well do compile_info = '-DCOMPILE_INFO=' + ........ followed by add_project_arguments(compile_info, language: 'cpp')

Or, you could do:

add_project_arguments('-DCOMPILE_INFO=' + ' '.join(compile_info), language : 'cpp')

Because the .join() should apply to the single-space separator token.

I tried this:

--- a/meson.build +++ b/meson.build @@ -203,24 +203,31 @@ libgerbera_deps = [ uuid_dep, ] -compile_info = [ '-DCOMPILE_INFO="' + - 'WITH_MAGIC=' + magic_dep.found().to_string().to_upper() + ' ' + - 'WITH_MYSQL=' + mysql_dep.found().to_string().to_upper() + ' ' + - 'WITH_CURL=' + curl_dep.found().to_string().to_upper() + ' ' + - 'WITH_INOTIFY=' + get_option('inotify').to_string().to_upper() + ' ' + - 'WITH_JS=' + duktape_dep.found().to_string().to_upper() + ' ' + - 'WITH_TAGLIB=' + taglib_dep.found().to_string().to_upper() + ' ' + - 'WITH_AVCODEC=' + avcodec_dep.found().to_string().to_upper() + ' ' + - 'WITH_FFMPEGTHUMBNAILER=' + ffmpegthumbnailer_dep.found().to_string().to_upper() + ' ' + - 'WITH_EXIF=' + exif_dep.found().to_string().to_upper() + ' ' + - 'WITH_EXIV2=' + exiv2_dep.found().to_string().to_upper() + ' ' + - 'WITH_SYSTEMD=' + get_option('systemd').to_string().to_upper() + ' ' + - 'WITH_LASTFM=' + lastfm_dep.found().to_string().to_upper() + ' ' + - 'WITH_DEBUG=' + get_option('debug').to_string().to_upper() + ' ' + - 'WITH_TESTS=' + get_option('tests').to_string().to_upper() + '"' -] - -add_project_arguments('-DCOMPILE_INFO='.join(compile_info), language : 'cpp') +compile_info = '-DCOMPILE_INFO="' +foreach info: [ + ['WITH_MAGIC', magic_dep.found()], + ['WITH_MYSQL', mysql_dep.found()], + ['WITH_CURL', curl_dep.found()], + ['WITH_INOTIFY', get_option('inotify')], + ['WITH_JS', duktape_dep.found()], + ['WITH_TAGLIB', taglib_dep.found()], + ['WITH_AVCODEC', avcodec_dep.found()], + ['WITH_FFMPEGTHUMBNAILER', ffmpegthumbnailer_dep.found()], + ['WITH_EXIF', exif_dep.found()], + ['WITH_EXIV2', exiv2_dep.found()], + ['WITH_SYSTEMD', get_option('systemd')], + ['WITH_LASTFM', lastfm_dep.found()], + ['WITH_DEBUG', get_option('debug')], + ['WITH_TESTS', get_option('tests')], + ] + + found = info[1].to_string().to_upper() + compile_info += 'WITH_@0@=@1@ '.format(info[0], found) +endforeach + +compile_info += '"' + +add_project_arguments(compile_info, language : 'cpp') # mangix hacks add_project_arguments('-DGIT_COMMIT_HASH=""', language : 'cpp')

works. Except the last entry ends in a space.

compile_info should be declared as a [] and += adds new array elements.

Or, you could do:

add_project_arguments('-DCOMPILE_INFO=' + ' '.join(compile_info), language : 'cpp')

Modern build system. Has the concept of wraps, which simplify dependencies when they are missing from the OS. Also has a clang-format ninja target, which can remove the separate CI for it. Signed-off-by: Rosen Penev <rosenp@gmail.com>

neheb force-pushed the meson branch 3 times, most recently from a000d3c to 5970ffd Compare September 2, 2021 04:02

neheb force-pushed the meson branch from 5970ffd to ecefb25 Compare September 6, 2021 06:25

eli-schwartz reviewed Sep 13, 2021

View reviewed changes

neheb force-pushed the meson branch from ecefb25 to b02d9a7 Compare September 13, 2021 06:02

neheb force-pushed the meson branch from b02d9a7 to ad7bb40 Compare September 13, 2021 20:05

eli-schwartz reviewed Sep 13, 2021

View reviewed changes

meson.build Outdated Show resolved Hide resolved

neheb force-pushed the meson branch 2 times, most recently from 24648dd to 55facea Compare September 14, 2021 03:52

neheb force-pushed the meson branch 2 times, most recently from ef2de62 to 58d4a67 Compare September 14, 2021 04:04

eli-schwartz reviewed Sep 14, 2021

View reviewed changes

neheb force-pushed the meson branch 2 times, most recently from 262571e to e1ed30e Compare October 6, 2021 04:29

neheb force-pushed the meson branch 2 times, most recently from bccf895 to c978b99 Compare November 19, 2021 04:10

neheb force-pushed the meson branch from 003c1d6 to 8cf67c9 Compare May 4, 2022 00:51

neheb force-pushed the meson branch from 8cf67c9 to f53ec24 Compare May 13, 2022 23:17

neheb force-pushed the meson branch 2 times, most recently from 07bc610 to 00c7fd5 Compare May 25, 2022 00:31

neheb force-pushed the meson branch 11 times, most recently from 332d0c4 to 2e8f4d5 Compare June 16, 2022 20:23

neheb force-pushed the meson branch 2 times, most recently from 8e02c8d to 3d9521f Compare June 25, 2022 05:55

neheb force-pushed the meson branch from 3d9521f to b1f761a Compare July 9, 2022 07:26

neheb force-pushed the meson branch 7 times, most recently from 16d72c6 to fcb8073 Compare January 26, 2024 01:37

add meson build system

ff52856

Modern build system. Has the concept of wraps, which simplify dependencies when they are missing from the OS. Also has a clang-format ninja target, which can remove the separate CI for it. Signed-off-by: Rosen Penev <rosenp@gmail.com>

neheb force-pushed the meson branch 2 times, most recently from ef7f2b8 to fbba2fb Compare January 26, 2024 02:07

meson CI

fe95d47

neheb force-pushed the meson branch from fbba2fb to fe95d47 Compare January 26, 2024 02:08

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add meson #1868

add meson #1868

neheb commented Aug 30, 2021 •

edited

Loading

whyman commented Aug 30, 2021

neheb commented Sep 2, 2021

KarlStraussberger commented Sep 2, 2021

neheb commented Sep 6, 2021 •

edited

Loading

eli-schwartz left a comment

eli-schwartz Sep 13, 2021

neheb Sep 13, 2021

neheb commented Sep 13, 2021

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021 •

edited

Loading

neheb commented Sep 13, 2021 •

edited

Loading

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021

neheb commented Sep 14, 2021

eli-schwartz Sep 14, 2021

neheb Sep 14, 2021

neheb Sep 14, 2021

eli-schwartz Sep 14, 2021

neheb Sep 14, 2021

eli-schwartz Sep 14, 2021

add meson #1868

Are you sure you want to change the base?

add meson #1868

Conversation

neheb commented Aug 30, 2021 • edited Loading

whyman commented Aug 30, 2021

neheb commented Sep 2, 2021

KarlStraussberger commented Sep 2, 2021

neheb commented Sep 6, 2021 • edited Loading

eli-schwartz left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

neheb commented Sep 13, 2021

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021 • edited Loading

neheb commented Sep 13, 2021 • edited Loading

neheb commented Sep 13, 2021

eli-schwartz commented Sep 13, 2021

neheb commented Sep 14, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

neheb commented Aug 30, 2021 •

edited

Loading

neheb commented Sep 6, 2021 •

edited

Loading

eli-schwartz commented Sep 13, 2021 •

edited

Loading

neheb commented Sep 13, 2021 •

edited

Loading