Opened 8 years ago

Closed 8 years ago

#8148 closed enhancement (fixed)

libreoffice-5.2.0.4

Reported by: bdubbs@… Owned by: ken@…
Priority: normal Milestone: 7.10
Component: BOOK Version: SVN
Severity: normal Keywords:
Cc:

Description

New version

Change History (15)

comment:1 by Pierre Labastie, 8 years ago

Owner: changed from blfs-book@… to Pierre Labastie
Status: newassigned

comment:2 by ken@…, 8 years ago

If it's any help, I can take this one. Seems to work ok. It no longer recognizes

--with-system-npapi-headers

in reply to:  2 ; comment:3 by Pierre Labastie, 8 years ago

Owner: Pierre Labastie removed
Status: assignednew

Replying to ken@…:

If it's any help, I can take this one. Seems to work ok. It no longer recognizes

--with-system-npapi-headers

OK, go ahead. If the above switch is not recognised, does it mean that we should not recomment npapi-sdk anymore?

in reply to:  3 comment:4 by Pierre Labastie, 8 years ago

Replying to pierre.labastie:

Forgot to say(write) "thanks"! Updating packages in the 50-100 SBU range is not hard, but long!

in reply to:  3 comment:5 by ken@…, 8 years ago

Owner: set to ken@…
Status: newassigned

Replying to pierre.labastie:

Replying to ken@…:

If it's any help, I can take this one. Seems to work ok. It no longer recognizes

--with-system-npapi-headers

OK, go ahead. If the above switch is not recognised, does it mean that we should not recomment npapi-sdk anymore?

Yes. There is no reference to npapi in my logs, and configure failed until I removed that switch.

comment:6 by ken@…, 8 years ago

Hmm, built it yesterday with -j8 and all was fine. Today, with -j4 it fails a unit test (vcl_timer.test). Google led me to

https://lists.freedesktop.org/archives/libreoffice/2016-February/073342.html

All I had installed after libreoffice was texlive, my AV stack, and then I'd attempted to use one texlive otf font (although I think a couple more got pulled in). So, based on that response I removed the conf file I'd added. I then had to drop a couple of other settings in my .Xresources - first, the 100x35 urxvt screen used the whole width om my monitor instead of about 60%, and then it was too tiny to read. But now my term looks much as it always used to, and when I resume 'make build' it fails again at that test.

I suppose a fresh build might get through, but I don't feel very confident.

comment:7 by ken@…, 8 years ago

Didn't help. I then went back to the buildscript I used yesterday [ with -j8 and no reliable measurements ], altered that to build for /opt/libreoffice and ran it with -j4 : that too has just failed. All the failures are in the vcl_timer.test.

comment:8 by bdubbs@…, 8 years ago

Can you just remove or disable the test?

comment:9 by Pierre Labastie, 8 years ago

As far as I understand, the segfault mentionned in the thread above is supposed to be fixed in 5.2. Anyway, I'll try to build libreoffice today (building boost right now).

comment:10 by Pierre Labastie, 8 years ago

Libreoffice built OK on my machine, with book instructions (except removing the --with-system-npapi switch). FWIW: 36 SBU at -j5 8GB (594 MB installed)

comment:11 by ken@…, 8 years ago

Pierre, thanks for those notes.

I was mistaken when I said that a couple more texlive fonts got pulled in. What actually got installed by LO were a load of fonts in /usr/lib/libreoffice/share/fonts/truetype/ : Caladea, Carlito. DejaVu*, Liberation*, Lin*, OpenSans, PT_Serif, Source*. Of those, I already had DejaVu*, Liberation, Lin (actually, it installed the older Lin*G).

And the texlive font I had made available was SourceCodePro.

I'll look again at the switches, and perhaps delete my initial LO installation.

comment:12 by ken@…, 8 years ago

I suppose I got "lucky" the first time - all subsequent builds failed at the same place, even after I removed the installed files.

I managed to disable that test, but this is a case where a patch would be more obvious about what is being done. A small hunk needs to be removed - I was going to c hange the 'ifeq' which wraps the test to 'ifneq', but that also disables the main hunk of tests in this particular file. FWIW,

sed -i '113,117d' vcl/Module_vcl.mk

And since we still claim "The instructions below will only build the package without running any unit tests. If you prefer to run the unit tests, replace make build with make" I'm going to look at turning off the unit tests in their entirety.

comment:13 by ken@…, 8 years ago

For building without the unit tests:

make build-check

Looking at fedora's patch to disable the firebird unit test at

http://pkgs.fedoraproject.org/cgit/rpms/libreoffice.git/plain/0001-disable-firebird-unit-test.patch

I guess that a sed like

sed -i '/CppunitTest_vcl_timer/d' vcl/Module_vcl.mk

would do the right thing. But unit tests have caused a lot of pain in the last couple of years, and upstream regard our builds as "unsupported configuration" (from memory), so better to just drop the tests.

comment:14 by ken@…, 8 years ago

Fixed at r17714.

comment:15 by Douglas R. Reno, 8 years ago

Resolution: fixed
Status: assignedclosed

Did you forget to close the ticket, or are you still working on it?

Note: See TracTickets for help on using tickets.