On 07/08/2012, at 9:28 AM, Derek Atkins wrote: > Have the script do "printenv > /tmp/script.env.$$" and then compare the > environments? > > -derek > > On Tue, August 7, 2012 11:57 am, John Ralls wrote: >> >> On Aug 7, 2012, at 7:53 AM, Geert Janssens wrote: >>> At least I'm giving you plenty of opportunity to test ;) I've got >>> another set of patches waiting for the sync to happen before I can svn >>> dcommit them. >>> >>> I hope you can sort this out soon. If there's anything I can do to help, >>> just let me know. >> >> I just ran it manually, so you should be able to git-update now. >> >> I can actually test locally using nc. The problem appears to be a subtle >> difference (not uid or gid) between the environment created by xinetd and >> my regular shell. If the script is launched by xinetd it fails, but if I >> run it from a regular shell it works.
Wasn't all that subtle after all, it was $PATH. For reasons known only to the gods, between 20:31 July 15th (the last successful run) and 00:31 on the 16th (the first failure), xinetd decided to stop loading the PATH set in my environment when executing the script and so got the distro-installed version of git. I've added an explicit PATH setting to the script and now it works from nc again. Regards, John Ralls _______________________________________________ gnucash-devel mailing list [email protected] https://lists.gnucash.org/mailman/listinfo/gnucash-devel
