Re: New Unit Test for tdf#130960 causing Build Failure on i686

On Fri, May 15, 2020 at 10:37:20AM +0200, Xisco Fauli wrote:
> If I remember well, this is the same problem Rene mentioned in IRC one
> day, which is related to the timezone.

Yes, it is.

> @Rene, do you have any update on this ?

I pointed that issue out at April, 17:

12:43 < _rene_> uhm.
12:43 < _rene_> - Row inserted  Sheet1.1:1              04/05/2020
15:01:10      (Row 1:1 inserted)
12:43 < _rene_> ?                                                   ^
12:43 < _rene_> + Row inserted  Sheet1.1:1              04/05/2020
17:01:10      (Row 1:1 inserted)
12:43 < _rene_> ? 
12:43 < _rene_> does the UITest assume TZ=UTC?

(no reaction).

Then it was discussed again at April, 24:

15:47 < _rene_> x1sc0:
http://www.rene-engelhard.de/~rene/libreoffice_7.0.0~alpha1~git20200423-1_amd64.build
[...]
16:09 <@x1sc0> _rene_, it's really weird you get 15:01:10 there, see
what I get when I open sc/qa/uitest/calc_tests/data/tdf131907.odt ->
https://imgur.com/a/Yb4357m
[...]
quit [Quit: Leaving]
16:10 <@x1sc0> _rene_, the test was added more than a week ago <
4a581f12e0d0fa105f053904300b729acb18194a >
16:11 < IZBot> core - tdf#130960: Add UItest -
https://git.libreoffice.org/core/commit/4a581f12e0d0fa105f053904300b729acb18194a
16:11 < _rene_> I know, failed since the
16:11 < _rene_> n
16:11 <@x1sc0> what happens if you open
sc/qa/uitest/calc_tests/data/tdf131907.odt go to Edit - Track Changes -
Manage ?
16:12 < loircbot> LibreOffice (online) michael.meeks *  (3 files in 2
dirs): Proxy: poll for output space if we need waking.
16:14 < _rene_> how old can the version be? :)
16:15 < _rene_> (and I have a sid VM here, which has timezone setting,
too, so it might not say much)
16:15 < _rene_> build is in a minimal chroot
16:16 < _rene_> .oO ( or I continue the build, ignoring make check and
test with that in said VM, but then the timezone thing still holds )
16:17 < _rene_> x1sc0: can try with a snapshot from April, 12...
[...]
joined #libreoffice-dev
16:20 < _rene_> x1sc0: a LO 6.4.3 in a "normal" system shows 17:01 here
16:21 < _rene_> x1sc0: AHA. And TZ=UTC localc
sc/qa/uitest/calc_tests/data/tdf131907.odt ends up as 15:01
16:22 < _rene_> so it _does_ look TZ related
[...]
16:23 < _rene_> (even though sticking TZ=UTC in solenv/gbuild/UITest.mk
didn't work for some reason. tried.)
16:23 <@x1sc0> _rene_, i'm wondering whether is a bug, I believe the
change time should be the same despite of the timezone
16:24 < _rene_> then you'd need to tell the timezone there
16:24 < _rene_> (or make it UTC in all cases.)
16:25 < _rene_> I would accept that, but I am sure people out there
won't....
16:25 < _rene_> (especially the frenchies)
16:27 < _rene_> in any case the build has to work in a chroot without
any TZ set ;), well, actually:
16:27 < _rene_> # cat /etc/timezone 
16:27 < _rene_> Etc/UTC
16:27 < _rene_> maybe it's TZ=Etc/UTC
16:28 < _rene_> (but then why did it work in my "normal" system....)
16:28 < _rene_> (TZ=UTC that is)
[...]
16:43 <@erAck> Fwiw, yes, the timestamp in track changes *is* timezone
aware and has to be (consider editing a shared document on a sever with
clients in different timezones); stored in document in UTC
16:43 <@erAck> so a UI test not considering that will fail
16:44 < loircbot> LibreOffice (online) michael.meeks *  (net/Socket.cpp
wsd/ProxyProtocol.cpp): Proxy: ensure dumpState dumps via the
MessageHandlerInterface too.
16:44 < _rene_> so it's a bug in the test...
16:44 < _rene_> it should force UTC

That I tried:

https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/-/commit/01441439ee17e6c63cb6e08449e4934a9bfa2aa1
https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/-/commit/8467ebd7f96e40d506a176e995bcb77112286b71

But as I said the next day:

09:46 < _rene_> erAck_[t], x1sc0: hmm, and now that exact test hangs...

where I just gave up for alpha1[1], hoping you would care about your own
test and fix it ;-)

So no: no news from me except the known stuff above.

Now to the mail :-):

> > ======================================================================
> > ERROR: test_tdf66263_Protect_Records (trackedChanges.CalcTrackedChanges)
> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >   File "/core/sc/qa/uitest/calc_tests4/trackedChanges.py", line 51, in 
> > test_tdf66263_Protect_Records
> >     gridwin = xCalcDoc.getChild("grid_window")
> > uno.com.sun.star.uno.RuntimeException: Could not find child with id: grid_window

No idea whether that (and similar otherwise) is affected by TZ. Is this
two bugs?

> > ----------------------------------------------------------------------
> > Traceback (most recent call last):
> >   File "/core/sc/qa/uitest/calc_tests4/trackedChanges.py", line 33, in test_tdf131907
> >     self.assertEqual(get_state_as_dict(xChangesList.getChild('0'))["Text"], text)
> > AssertionError: 'Row inserted \tSheet1.1:1\t \t04/05/2020 11:01:10\t (Row 1:1 inserted)' != 
> > 'Row inserted \tSheet1.1:1\t \t04/05/2020 17:01:10\t (Row 1:1 inserted)'
> > - Row inserted      Sheet1.1:1              04/05/2020 11:01:10      (Row 1:1 inserted)
> > ?                                                       ^
> > + Row inserted      Sheet1.1:1              04/05/2020 17:01:10      (Row 1:1 inserted)
> > ?                                                       ^

Mine is Europe/Berlin and as shown above I get

- Row inserted  Sheet1.1:1              04/05/2020 15:01:10      (Row 1:1 inserted)
?                                                   ^
+ Row inserted  Sheet1.1:1              04/05/2020 17:01:10      (Row 1:1 inserted)

(which incidentially completely matches Europe/Berlin. (UTC+2)).

Regards,

Rene

[1] sticking a timeout into the build scripts and because that causes an error, ignoring
test errors. It's alpha over all, but there's a beta upcoming...