Thu 26 Nov 2015 05:09:11 AM UTC, comment #10:
No progress upstream in 2 years. Closing report.
|
Sun 28 Dec 2014 04:20:16 PM UTC, comment #9:
What do we want to do with this bug? The problem is upstream in gnulib. Have they been contacted about it as suggested in comment #4 from Carlo?
|
Tue 29 Oct 2013 07:02:05 AM UTC, comment #8:
Yes, the build does fail on OSX if '/usr/bin/sed' shiped b Applle is used.
I haad to do
to make sure the sed installed by macports is used.
|
Tue 29 Oct 2013 01:22:46 AM UTC, comment #7:
Carlo, I agree that gnulib is calling sed instead of $SED, but I notice you didn't say whether the build fails in the libgnu directory or not with OSX's "broken" sed.
Automake also puts calls to sed, not $SED, in the generated Makefiles.
|
Mon 21 Oct 2013 11:29:43 PM UTC, comment #6:
However, gnulib is special because once the bug is fixed in gnulib, we need a reminder to update our hg mirror of gnulib and the subrepo pointer in order to get the fix into Octave. I think leaving this report open on our side serves as that reminder.
|
Wed 16 Oct 2013 02:49:49 PM UTC, comment #5:
What I had in mind is that a report would be filed with the other bug trackers, say for gnulib, since there isn't much we can do at this end.
When we find errors that are attributable to gnuplot, for example, we generally close the bug here. The bug report can always be found by searching on keywords and so someone in the future who encounters the same problem can discover where to go without having to leave the report open.
|
Wed 16 Oct 2013 09:04:35 AM UTC, comment #4:
Rik,
Thanks for the partial fix!
I'll contact the developers of
other projects to try get a fix
there too.
In the meanwhile, please don't
close the bug, the fact that this
issue is to be fixed elsewhere does
not mean it's solved yet.
|
Tue 15 Oct 2013 06:14:04 PM UTC, comment #3:
Carlo, I pushed a changeset that uses $SED and $GREP wherever it is possble (http://hg.savannah.gnu.org/hgweb/octave/rev/1978a6c76aa9). Unfortunately, it seems that there are a lot of bare instances of 'sed' coming from either gnulib or autoconf (check the build-aux directory for instance). I don't see an easy way to get rid of those. I'm tempted to close the bug report as the remaining problems are not Octave, but other project's bugs.
|
Tue 15 Oct 2013 06:50:11 AM UTC, comment #2:
Yes,
There seem to be other occurrences of use of 'sed' not related to gnulib, for example
doing
I see at least those:
Do you need me to post the complete config.log and make.log?
thanks!
c.
|
Tue 15 Oct 2013 06:14:06 AM UTC, comment #1:
I think the output you quote is from gnulib, so maybe we should bring up this issue with the gnulib maintainers?
Are there any instances of sed used once the build moves out of the libgnu directory? If so, then we can fix those.
|
Tue 15 Oct 2013 05:48:00 AM UTC, original submission:
I am building Octave 3.7.7 on OSX 10.8.5
which provides a broken version of sed in /usr/bin.
I installed GNU sed as /opt/local/bin/gsed.
If I configure with
GNU sed is correctly detected by configure:
yet, when I run 'make', 'sed' is still used:
this is the first occurence of sed in the output of make but there are many more down the way.
-verbatim-
|