Travis build broken with commit 8d904c1209d3973f5f2bbf948c0c88c436dbdebc

[Geert Janssens]

On Friday 18 December 2015 13:54:19 John Ralls wrote:
> > On Dec 18, 2015, at 9:53 AM, Geert Janssens
> > <geert.gnucash at kobaltwit.be> wrote:
> > 
> > That's "Correct detection of marker commodity splits."
> > 
> > It's more likely though it's the commit right before it that breaks
> > the travis test run:
> > 
> > Use @SHELL@ instead of hardcode /bin/sh in test shell scripts.
> > 
> > FYI the log of the test run is here:
> > https://s3.amazonaws.com/archive.travis-ci.org/jobs/97579062/log.txt
> > 
> > That's the raw version because the interactive version was just too
> > large and froze my firefox.
> 
> Man, that’s noisy.

True. In interactive view you normally only display the sections of interest.

> 
> Anyway, the cause is most likely  9f3d6f, Bug 759224 - illegal dates
> in stock transactions cause corrupt file, which removes the
> dependency on strptime and switches to using
> gnc_iso8601_to_timespec_gmt. These errors show up in the middle of
> the test log: make[6]: Entering directory
> `/home/travis/build/Gnucash/gnucash/src/backend/xml/test' FAILURE
> timespec_secs test-date-converting.c:67 not equal ints are 1804289383
> and 8362594656000 …
> 
> I tried on my Debian VM and the tests failed by crashing rather than
> returning the wrong result. I fixed it by making the same changes I’d
> made to master, disabling creating nanoseconds in test-engine-stuff.c
> and trying to parse them in sixtp-utils.c, and that corrected the
> make check failure on Debian. It also seems to have fixed it on
> Travis.
> 
> Regards,
> John Ralls

Great,

Thanks for the prompt response!

Geert