This is the mail archive of the
mailing list for the Cygwin project.
cygport: broken vs. autotools by "set -e"
- From: Dave Korn <dave dot korn dot cygwin at gmail dot com>
- To: Cygwin Mailing List <cygwin at cygwin dot com>
- Date: Thu, 19 Jan 2012 23:45:31 +0000
- Subject: cygport: broken vs. autotools by "set -e"
I just updated my Cygwin installation for the first time since Oct 26 last
year, and now I'm unable to successfully run cygport builds any more. The
builds fail because of the spontaneous exiting of one or other of the scripts
that cygport invokes - I've had autoconf-2.68 spontaneously exiting during the
autoreconf stage, and if I skip over that the configure scripts do the same thing.
On closer inspection, it turns out that they were exiting first time they
executed subcommands that failed. I added 'set -o' to my cygport script, both
inline so it would be executed when the script gets sourced, and inside my
override of src_compile so I could see what was happening just before the
configure call, and sure enough it showed me that errexit was off at the start
of the run but had been set on by the time execution reached src_compile.
Commenting-out the "set -e;" line at the start of /usr/bin/cygport has fixed
this problem, and my builds now run just fine, but huh? I checked in git;
that line has been there since like forever, so why is it giving me trouble
now? Is there something I could have changed in my environment or startup
scripts that is causing this -e to propagate to subshells that it didn't used
to, or did there used to be a mechanism in cygport that would have had the
effect of turning it off for subshells that has now been removed for some reason?
Problem reports: http://cygwin.com/problems.html
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple