qt-4.8.6

[Fernando de Oliveira]

​http://download.qt-project.org/official_releases/qt/4.8/4.8.6/qt-everywhere-opensource-src-4.8.6.tar.gz

​http://download.qt-project.org/official_releases/qt/4.8/4.8.6/changes-4.8.6

Qt 4.8.6 is a bug-fix release. It maintains both forward and backward
compatibility (source and binary) with Qt 4.8.0, 4.8.1, 4.8.2, 4.8.3,
4.8.4, and 4.8.5.
For more details, refer to the online documentation included in this
distribution. The documentation is also available online:

   http://qt-project.org/doc/qt-4.8/

...

[Fernando de Oliveira]

Program "designer" is segfaulting, After an strace, I am suspecting something is missing, because it is not finding many icons. And this alone could make desiner to sefault.

I cannot see how to love this until now.

Will attach the strace, in case somebody have time to help, please.

Thanks in advance.

[Fernando de Oliveira]

strace of designer run from lxterminal

[Fernando de Oliveira]

I've spent some more time with this.

qt-4.8.5 (current in the book) segfaults in the same way.

Not finding some (many) icons must not be the problem: in LFS-7.5, designer qt-4.8.5 works perfectly.

I never used this program other than testing that qt is working.

In LFS-svn, where it is problematic, diferences that I remember: systemd is installed, ICU is newer, gcc-4.9.0.

[Armin K]

As I mentioned in one of LFS tickets, Qt4 also segfaults here when built with gcc-4.9.0. However, building it without "-release" switch fixed the problem for me. It seems that when using -release, some flags are used which render some of the code unusable.

[Fernando de Oliveira]

Thank you very much, Armin! I had nightmares this night.

Did not remember.

I already had replaced release by debug, and it is being built at this moment. Hope that the effect is the same as yours.

[Armin K]

I've found out that this gcc commit might fix it:

​http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=210079

Note that it requires rebuilding GCC. I am building early gcc 4.9.1 snapshot and will report back if the error goes away.

[Fernando de Oliveira]

Thanks, Armin.

As I write below, I knew about the patch, but was not considering to update gcc. Thought that was the "position" of LFS, according to an earlier discussion, there. But I agree with you: if we have something that fixes this and other problems, why not use immediately?

As soon I have your reply, will start, then, patching gcc.

Yes, I was sending what follows, but some tools update in FF was giving me LFS off line, only when I received your post, I noticed that the site should be on. Following is waht I wrote before your post.

Build with debug failed.

Built without release did not solve designer crash.

Only info from gdb is a warning "could not load shared library symbols for linux.gate.so.1".

Upstream is already aware, but seems will not fix: gcc-4.9.0 bug, not qt.

​https://bugreports.qt-project.org/browse/QTBUG-38733

A simple patch to gcc would solve the problem:

"This seems to be ​http://gcc.gnu.org/PR60965. The patch in the comments there avoids the crash (after only recompiling JSGlobalData.cpp with the patched compiler)."

This is also reproduced at redhat:

​https://bugzilla.redhat.com/show_bug.cgi?id=1091482

[bdubbs@…]

If you put the patch in the patches repo, I'll add it to LFS today. I don't know if it's needed in Chapter 5, but the conservative thing to do is to add it to both gcc builds there too.

[Fernando de Oliveira]

OK, thanks. Just waiting Armin, so I can start my gcc rebuild.

[Armin K]

It seems to work. I have built gcc snapshot which includes already mentioned patch and rebuilt Qt4 afterwards. KDE now starts up just fine, whereas it previously wouldn't even show the login screen.

[Fernando de Oliveira]

Replying to Krejzi:

I've found out that this gcc commit might fix it:

​http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=210079

Note that it requires rebuilding GCC. I am building early gcc 4.9.1 snapshot and will report back if the error goes away.

I tried to do a patch from that address, but failed.

Seems what is there considers other previous commit?

In order to keep updating the book, a snapshot could be considered as the one in

​http://gcc.petsads.us/snapshots/4.9.0-RC-20140411/gcc-4.9.0-RC-20140411.tar.bz2

?

Really don't know what to do.

[Armin K]

That is gcc-4.9.0 rc and thus is older than gcc-4.9.0.

I have used "gcc version 4.9.1 20140507 (prerelease)"

Bruce says he added the patch, but I didn't see any commit to patches repo. Bruce?

[bdubbs@…]

I asked for someone working on this directly to post a patch. I haven't been working this today.

[Armin K]

I believe I could create a patch if I just could manage to "download" a patch from the link I posted above. Why on earth are people still stuck with cvs :|

[Fernando de Oliveira]

Replying to Krejzi:

That is gcc-4.9.0 rc and thus is older than gcc-4.9.0.

I have used "gcc version 4.9.1 20140507 (prerelease)"

Of course. I think it is because it was below:

[DIR] 4.9-20140507/ 08-May-2014 00:36 - [DIR] 4.9.0-RC-20140411/ 11-Apr-2014 17:23 -

that tricked my mind.

Bruce says he added the patch, but I didn't see any commit to patches repo. Bruce?

I saw Bruce already peplied this part.

[Fernando de Oliveira]

Replying to Krejzi:

I believe I could create a patch if I just could manage to "download" a patch from the link I posted above. Why on earth are people still stuck with cvs :|

There are four entries there. One for code, two for changelogs and one for tests code.

What I did, first: copied pasted the "code patch" but one chunk was rejected.

Next, I downloaded the full file, diffed from the one in 4.9.0 and tried to compile the patched source code (equivalent to the original with a replaced file), but make failed.

It was not a surprise, because the first method I used indicated that other previous modification after 4.9.0 had already been made to that file.

[Armin K]

​http://gcc.gnu.org/git/?p=gcc.git;a=patch;h=6977ccded4335bd8c0215ccfcab861598f8e9388

This is the patch that would need to be added. I didn't test the build. Do note that patches from LFS Bug about FLAC or whatever are not part of this.

[Armin K]

I have created a patch that contains previously mentioned one as well as one mentioned in LFS bug and added it to the patches repository:

​http://www.linuxfromscratch.org/patches/downloads/gcc/gcc-4.9.0-upstream_fixes-1.patch

I have verified that gcc builds fine with current LFS instructions. I didn't run the testsuite.

[Fernando de Oliveira]

Thank you very much indeed!

Will finish one more package I have started and will go back to this one.

One small thing, I don't know if it will bother you or not:

Initial Package Version: 4,9,9.

My old bad eyes see commas, not dots, but I think that everybody would understand. But as I know you are perfectionist...

But again, this patch is much appreciated!!!

[Fernando de Oliveira]

Replying to fo:

Initial Package Version: 4,9,9.

Oh! version is 4.9.0, not 4.9.9, isn't it?

[Armin K]

That's what happens when you're in a rush. Should be fixed now. Thanks!

[Fernando de Oliveira]

Worked perfectly. Thanks again.

Bruce, please, remember to add to LFS. I did not run the tests.

Fixed a r13029.