at-spi2-core @2.28.0 does not build on PPC Tiger (ld: -rpath can only be used when targeting Mac OS X 10.5 or later)

[ballapete (Peter "Pete" Dyballa)]

Compared to bug #56932 now glib2 @2.56.2_0+x11 (active) platform='darwin 8' archs='ppc' date='2018-08-25T00:34:01+0200' instead of glib2 @2.56.1_0+x11.

[48/53] Compiling C object 'registryd/registryd@@at-spi2-registryd@exe/ucs2keysym.c.o'.
[49/53] Linking target registryd/at-spi2-registryd.
FAILED: registryd/at-spi2-registryd 
/opt/local/bin/gcc-apple-4.2  -o registryd/at-spi2-registryd 'registryd/registryd@@at-spi2-registryd@exe/de-marshaller.c.o' 'registryd/registryd@@at-spi2-registryd@exe/registry-main.c.o' 'registryd/registryd@@at-spi2-registryd@exe/registry.c.o' 'registryd/registryd@@at-spi2-registryd@exe/introspection.c.o' 'registryd/registryd@@at-spi2-registryd@exe/deviceeventcontroller.c.o' 'registryd/registryd@@at-spi2-registryd@exe/reentrant-list.c.o' 'registryd/registryd@@at-spi2-registryd@exe/deviceeventcontroller-x11.c.o' 'registryd/registryd@@at-spi2-registryd@exe/display.c.o' 'registryd/registryd@@at-spi2-registryd@exe/event-source.c.o' 'registryd/registryd@@at-spi2-registryd@exe/ucs2keysym.c.o' -L/opt/local/lib -Wl,-dead_strip_dylibs -pipe -Os -arch ppc -Wl,-headerpad_max_install_names -arch ppc atspi/libatspi.0.dylib /opt/local/lib/libgio-2.0.dylib /opt/local/lib/libgobject-2.0.dylib /opt/local/lib/libglib-2.0.dylib /opt/local/lib/libintl.dylib -Wl,-framework -Wl,CoreFoundation /opt/local/lib/libdbus-1.dylib -Wl,-framework -Wl,CoreFoundation -ldl /opt/local/lib/libX11.dylib /opt/local/lib/libXtst.dylib /opt/local/lib/libXi.dylib '-Wl,-rpath,$ORIGIN/../atspi'  
ld: -rpath can only be used when targeting Mac OS X 10.5 or later
collect2: ld returned 1 exit status
[50/53] Generating Atspi-2.0.gir with a custom command.
at-spi2-core-2.28.0/atspi/atspi-gmain.c:44: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< the parent GSource */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:45: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< the connection to dispatch */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:98: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< the main context */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:99: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< all IOHandler */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:100: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< all TimeoutHandler */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:101: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< NULL if this is really for a server not a connection */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:102: Error: Atspi: Skipping invalid GTK-Doc comment block:
/**< DBusGMessageQueue */
^
at-spi2-core-2.28.0/atspi/atspi-gmain.c:513: Error: Atspi: Skipping invalid GTK-Doc comment block:
/** @} */
^
g-ir-scanner: link: /opt/local/bin/gcc-apple-4.2 -arch ppc -o /opt/local/var/macports/build/_opt_local_var_macports_sources_lil.fr.rsync.macports.org_release_tarballs_ports_gnome_at-spi2-core/at-spi2-core/work/build/tmp-introspectzUfIsu/Atspi-2.0 /opt/local/var/macports/build/_opt_local_var_macports_sources_lil.fr.rsync.macports.org_release_tarballs_ports_gnome_at-spi2-core/at-spi2-core/work/build/tmp-introspectzUfIsu/Atspi-2.0.o -L. -latspi -ldbus-1 -lgobject-2.0 -lglib-2.0 -lintl -lX11 -lXtst -lXi -L/opt/local/lib -lgio-2.0 -lgobject-2.0 -lgmodule-2.0 -lglib-2.0 -lintl -Wl,-framework -Wl,CoreFoundation
ninja: build stopped: subcommand failed.

[ballapete (Peter "Pete" Dyballa)]

Main.log from build with apple-gcc42 @5666.3_15 (active) platform='darwin 8' archs='ppc' date='2018-08-20T17:19:16+0200'

[dbevans (David B. Evans)]

The "Skipping invalid GTK-Doc comment block" errors are really warnings emitted by g-ir-scanner not gtk-doc in this instance. They are not fatal and only indicate that the scanner is skipping these improperly formatted generic comments which are erroneously using markers reserved for gtk-doc only. I see these same warnings in the log of a successful build on Mojave.

The log file lines that are relevant to this build failure seem to be these

246	:info:build [49/53] Linking target registryd/at-spi2-registryd.
247	:info:build FAILED: registryd/at-spi2-registryd
248	:info:build /opt/local/bin/gcc-apple-4.2  -o registryd/at-spi2-registryd 'registryd/registryd@@at-spi2-registryd@exe/de-marshaller.c.o' 'registryd/registryd@@at-spi2-registryd@exe/registry-main.c.o' 'registryd/registryd@@at-spi2-registryd@exe/registry.c.o' 'registryd/registryd@@at-spi2-registryd@exe/introspection.c.o' 'registryd/registryd@@at-spi2-registryd@exe/deviceeventcontroller.c.o' 'registryd/registryd@@at-spi2-registryd@exe/reentrant-list.c.o' 'registryd/registryd@@at-spi2-registryd@exe/deviceeventcontroller-x11.c.o' 'registryd/registryd@@at-spi2-registryd@exe/display.c.o' 'registryd/registryd@@at-spi2-registryd@exe/event-source.c.o' 'registryd/registryd@@at-spi2-registryd@exe/ucs2keysym.c.o' -L/opt/local/lib -Wl,-dead_strip_dylibs -pipe -Os -arch ppc -Wl,-headerpad_max_install_names -arch ppc atspi/libatspi.0.dylib /opt/local/lib/libgio-2.0.dylib /opt/local/lib/libgobject-2.0.dylib /opt/local/lib/libglib-2.0.dylib /opt/local/lib/libintl.dylib -Wl,-framework -Wl,CoreFoundation /opt/local/lib/libdbus-1.dylib -Wl,-framework -Wl,CoreFoundation -ldl /opt/local/lib/libX11.dylib /opt/local/lib/libXtst.dylib /opt/local/lib/libXi.dylib '-Wl,-rpath,$ORIGIN/../atspi' 
249	:info:build ld: -rpath can only be used when targeting Mac OS X 10.5 or later
250	:info:build collect2: ld returned 1 exit status

As you can see, the -rpath option is not supported by the version of ld supplied by Tiger. The solution is probably to use a newer version of ld if one is available or to modify this linker invocation to avoid the use of -rpath. Since I don't have a Tiger machine available to play with I'm afraid I can't be much further help. Reassigning this ticket to @kencu since this is really his area of expertise.

BTW, there's an upstream patch that quiets the gtk-doc error messages. I'll gp ahead and apply that to hopefully reduce the noise level in this case.

[dbevans (David B. Evans)]

In e2f7559308947878248a2c9eeb5fe65612534eb0/macports-ports (master):

at-spi2-core: patch to remove spurious gtk-doc markers

See #57030

[kencu (Ken)]

see 57975 where there are some thoughts on how to manage this issue.

adding rpath support to Tiger is likely not possible, as I believe it requires changes deep in the OS to do that.

But having ld64 ignore the rpath command is something that is pretty easy and might fix it for us, as macports does not use the rpath functions anyway, and (I believe) always installs executables and libraries with the full pathspecs on final installation.

[kencu (Ken)]

see also 56932 which is a duplicate of this, but with useful information.

[kencu (Ken)]

I don't plan on fixing this any time soon, and will reopen it if I get around to doing so.