gobject-introspection @1.60.2_5 fails to configure (with fixes)

[dsteck]

Configure failed in two ways. Tested on macports with everything uninstalled, Big Sur 11.2.3 (20D91), CLI tools 12.4.0.0.1.1610135815, as x86_64 (everything current as of now, 10 March 2021).

1. configure complains that it can't find gtkdocize

(see gobject-introspection-fail1.log)

solution: add port gtk-doc as a dependency

2. configure complains that it can't find Python.h; the issue is that CPP='', so the test is run without specifying a preprocessor

(see gobject-introspection-fail2.log)

solution: add two lines to portfile (not sure if this is the right fix, but it's how I was able to make it work):
configure.cpp 'gcc -E'
configure.post_args CPP=${configure.cpp}

[dsteck]

config.log for first problem

[dsteck]

config.log for second problem

[kencu (Ken)]

autoconf was updated a few days ago after about 8 years, and adds the gtkdocize problem.

We may see 1000 ports that need gtk-doc added as a build dep. Or perhaps we sort out how to add it to autoconf directly, but unfortunately at present there is a circular dependency issue (gtk-doc depends on glib2 which depends on autoconf which (maybe) depends on gtk-doc). There is another ticket tracking this issue.

For now you can activate the previous autoconf to get rid of this particular headache until more ports are fixed.

no idea about the cpp issue, but I would not be surprised if that also comes from the new autoconf.

We have only just begun to feel the pain coming from this autoconf upgrade I fear...

[jmroot (Joshua Root)]

No need to be dramatic, Ken. The change only affects ports that have a configure script that uses the GTK_DOC_CHECK macro, are running autoreconf, and specify the -i/--install option. There are only around 180 ports that could possibly be affected based on the latter two conditions. Of those, only the gnome-related ones are likely to actually be using GTK_DOC_CHECK.

If completely regenerating the build system is desired, adding the dependency is correct. If not, autoreconf -fvi isn't the command you're looking for, and either not using -i or only running a subset of the commands that autoreconf would run is the appropriate approach.

[dsteck]

Thanks for the info Ken and Joshua, and sorry to hear that a lot of maintenance work may lie ahead. That explains why I didn't have any trouble building this port a few weeks ago (I assumed it was related to the Xcode update). I'll leave it to people wiser than me to decide how to best deal with the issues here...

[kencu (Ken)]

Replying to jmroot:

No need to be dramatic, Ken.

Thanks, Josh. I guess we'll see, as time goes by.

[michaelld (Michael Dickens)]

I just bumped into this issue when working on ​https://trac.macports.org/ticket/62391 ... reverting the "autoconf" port did the trick to unblock for what I need here. Seems like some pretty specific conditions need to be met for the issue to be present, yet even 180 ports will induce some reasonable pain! I'm glad y'all are on top of this issue, and I hope it's fixed expeditiously :)

[jmroot (Joshua Root)]

You're likely in a better position to see if gobject-introspection works correctly after changes than I am, Michael—would you mind checking if it really needs a full autoreconf, or if use_autoconf plus use_automake is sufficient? If not, try autoreconf.args -fv (i.e. remove the -i)?

[jmroot (Joshua Root)]

Try this for the CPP issue:

configure.ac

@@ -48 +48 @@
 # Checks for programs.
 AC_PROG_CC
 AM_PROG_CC_C_O
+AC_PROG_CPP
 AC_PROG_MKDIR_P
 
 # Initialize libtool

[michaelld (Michael Dickens)]

@jmroot: this last change works nicely & makes a lot of sense. Seems like an oversight on the part of the upstream project: the function AM_CHECK_PYTHON_HEADERS calls a project-internal script m4/python.m4 where CPP is required in AC_TRY_CPP, but neither script actually checks for AC_PROG_CPP as you added there.

[jmroot (Joshua Root)]

In c8fd287b58e089ee06fb0e12aa141d8c7169cc44/macports-ports (master):

gobject-introspection: fix configure failure

Closes: #62428

[MarcusCalhoun-Lopez (Marcus Calhoun-Lopez)]

In fa4675525fb6dde4832fb9adc8419268af94858c/macports-ports (master):

seed: fix configure failure

For a similar problem, see #62428