gpsbabel 1.3.5 fails build - "configure.in:39: error: AC_SUBST"

[fracai]

gpsbabel 1.3.5 fails to build under Intel (iMac and MacBook) Mac OS X 10.5.4

output of the attempt:

--->  Fetching gpsbabel
--->  Verifying checksum(s) for gpsbabel
--->  Extracting gpsbabel
--->  Applying patches to gpsbabel
--->  Configuring gpsbabel
--->  Building gpsbabel with target all
Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_textproc_gpsbabel/work/gpsbabel-1.3.5" && make all " returned error 2
Command output: autoconf
configure.in:39: error: AC_SUBST: `' is not a valid shell variable name
configure.in:39: the top level
autom4te: /opt/local/bin/gm4 failed with exit status: 1
make: *** [configure] Error 1

Error: Unable to upgrade port: 1

[ryandesign (Ryan Carsten Schmidt)]

See same error message for fish in #16012.

[fracai]

A possible fix is listed ​here.

... edit the file config.in
to delete the lines mentioning the (obsolete) AC_EXEEXT.

I haven't yet tested this myself.

[ryandesign (Ryan Carsten Schmidt)]

Sorry, arno, I didn't see your message before committing a different change.

I can reproduce the problem with Mac OS X 10.5.4 and Xcode 3.1 on a MacBook Pro. I also have the autoconf 2.62 port installed.

The solution from #16012 isn't applicable here. There, we needed to patch configure.ac anyway because it couldn't properly detect iconv in Mac OS X. And configure.ac contained a definition that newer autoconf apparently doesn't like anymore so I fixed those definitions.

But here, I can't even see why autoconf would need to run. I mean I see why it is running: MacPorts runs "./configure" which generates Makefile from Makefile.in, and Makefile now has an mtime of "now" making it newer than the configure script. Then MacPorts runs "make all" and the Makefile has set up its dependencies such that the executable depends on the configure script, and the rule for building the configure script says to run autoconf. So even if I could fix the reported error, we'd be back to running "./configure", then "make all" which would cause "autoconf" to run and re-run the configure phase again. It's all nonsense. I nullified the rule for running autoconf to build the configure script since it's already built. Now the port builds for me.

Fix is in r38415. Thank you, arno, for reporting this problem.

Michael, you should report both of these problems (1. autoconf is run every time which shouldn't be necessary; 2. current version of autoconf complains) to the developers of gpsbabel when you get a chance.

[(none)]

Milestone Port Bugs deleted