librsvg @2.40.20_5: build fails on darwin 9, previous revision didn't

[kakuhen]

librsvg @2.40.20_4 is successfully installed on my Power Mac G5 running OS X 10.5.8

This current revision seems to have broken the vala build process; below is the relevant snippet from logs.

:info:build GdkPixbuf-2.0.gir:6382.55-6382.55: error: The symbol `GModule' could not be found
:info:build  6382 |         <type name="GModule.Module" c:type="GModule*"/>
:info:build       |                                                       ^
:info:build Generation failed: 1 error(s), 13 warning(s)

Attached to this ticket is the full main.log

[barracuda156]

Identical failure on Rosetta:

:info:build GdkPixbuf-2.0.gir:6382.55-6382.55: error: The symbol `GModule' could not be found
:info:build  6382 |         <type name="GModule.Module" c:type="GModule*"/>
:info:build       |                                                       ^
:info:build Generation failed: 1 error(s), 13 warning(s)
:info:build make[2]: *** [librsvg-2.0.vapi] Error 1
:info:build make[2]: Leaving directory `/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_graphics_librsvg/librsvg/work/librsvg-2.40.20'
:info:build make[1]: *** [all-recursive] Error 1

[barracuda156]

Maybe the header missing somewhere? ​https://developer.gimp.org/api/2.0/glib/glib-Dynamic-Loading-of-Modules.html

[barracuda156]

Replying to mascguy:

A similar error reported here: ​https://gitlab.gnome.org/GNOME/gdk-pixbuf/-/issues/193

But is also can be the case that vapigen is broken for PPC, which I hold is the case. Upstream ignores the problem: ​https://gitlab.gnome.org/GNOME/vala/-/issues/1297

[kencu (Ken)]

in your stacktrace, I notice vala is hanging just after calling posix_memalign… ​https://gitlab.gnome.org/GNOME/vala/-/issues/1297

systems < 10.6 don’t have posix_memalign, so I wrote a best-effort workalike using what Tiger and Leopard did have, and put it in legacysupport. It is not the same as what 10.6+ has, though, and it’s possible that is the source of this hang. You might look into the vala code and see if you can try a different workaround for posix_memalign instead.

the other thought is that it has something to do with trying to free the wrong memory, ie a variation on the libstdc++ “unaligned pointer being freed” issue we have not solved yet…

[kencu (Ken)]

upstream wanted you to run the vala test suite, I notice… without doing at least that, I doubt they will reply further. Nobody there will have an old powermac to play with.

[johnrosshunt]

I get an identical error building librsvg on 10.5 Intel. So, I tried the vala test suite and ALL the tests failed. That's because there's an upstream typo in build-aux/testrunner.sh. On line 193 change ';&' to ';;' and it will run properly. Under leopard about half the vala tests still failed. Building vala again with MacPorts clang-9.0 results in approximately 30-something tests failing -- most pertaining to DBUS. With over 1000 tests in the suite, that's pretty good. But, librsvg still wouldn't build for me. So, I looked at gdk-pixbuf2 and saw this:

$ sudo gdk-pixbuf-query-loaders --update-cache
gdk-pixbuf-query-loaders(89918) malloc: *** error for object 0xa0107240: Non-aligned pointer being freed (2)
*** set a breakpoint in malloc_error_break to debug
gdk-pixbuf-query-loaders(89918) malloc: *** error for object 0xa01071c0: Non-aligned pointer being freed (2)
*** set a breakpoint in malloc_error_break to debug

Could this be the problem?

[barracuda156]

Replying to johnrosshunt:

I get an identical error building librsvg on 10.5 Intel. So, I tried the vala test suite and ALL the tests failed. That's because there's an upstream typo in build-aux/testrunner.sh. On line 193 change ';&' to ';;' and it will run properly. Under leopard about half the vala tests still failed. Building vala again with MacPorts clang-9.0 results in approximately 30-something tests failing -- most pertaining to DBUS. With over 1000 tests in the suite, that's pretty good. But, librsvg still wouldn't build for me. So, I looked at gdk-pixbuf2 and saw this:

$ sudo gdk-pixbuf-query-loaders --update-cache
gdk-pixbuf-query-loaders(89918) malloc: *** error for object 0xa0107240: Non-aligned pointer being freed (2)
*** set a breakpoint in malloc_error_break to debug
gdk-pixbuf-query-loaders(89918) malloc: *** error for object 0xa01071c0: Non-aligned pointer being freed (2)
*** set a breakpoint in malloc_error_break to debug

Could this be the problem?

This also happens on PPC: #66159

[kencu (Ken)]

​https://gitlab.gnome.org/GNOME/librsvg/-/commit/104fb592b0c32f9e9e3e9246210682dfa0ce4a60

[kencu (Ken)]

In d50c5142902af4c3fcad0868446a185b33f2b40a/macports-ports (master):

librsvg: fix build on older systems

fallback librsvg had a missing vapi dep
closes: #65407

[kencu (Ken)]

the gdk-pixbuf2 malloc errors are handled here:

​https://github.com/macports/macports-ports/pull/16901