icedtea-2.5.0

[Fernando de Oliveira]

​http://icedtea.wildebeest.org/download/source/icedtea-2.5.0.tar.xz

[Fernando de Oliveira]

I don't have time: FIFA World Cup, Four matches just today.

[Pierre Labastie]

I am not much interested by FIFA cup ;-), but I have other things to do. Hope to be able to update by the end of the week.

[Fernando de Oliveira]

It is not only that. I tend to get less interested by packages that became or will become too different from what they were. just for the sake of the modification or the dev preference, and use my own scripts for them.

If it isn't broken, don't fix it. This is a very anti-narcissistic rule ;-)

New packages sometimes do interest me.

We have many problems that could be used for BLFS work time. LibreOffice update, that I could not make work in i686 is one of them.

[Pierre Labastie]

This update might take longer than anticipated. The patches do not apply anymore, and I have to regenerate them. Then, I'll have to build a 32 bit VM with the new LFS layout for generating the 32 bit binary. Sorry for the delay

[Fernando de Oliveira]

If it can help, DJ once suggested that the cacerts patch could be removed, after I included the instructions to generate them, explicitly, in Configurations.

It would be one less patch to worry about.

[Fernando de Oliveira]

Pierre, now you will have a 32 bit machine.

Please, I am stuck with LibreOffice-4.2.4.2 in i686. Would you mind to check if it a general problem and if you could help us (Bruce is already helping me) solving it?

[Pierre Labastie]

Thanks for the info about the cacert patch. For LO, I have seen your SOS, and I'd be happy to help, but it is a big beast I have never built before (neither on 32 nor on 64 bit), so I guess I'll first have to learn it. Furthermore, if it is a GCC error, it may be tricky to fix... Anyway, let me finish OJDK first.

[Pierre Labastie]

I am not very successful with tests. On 32 bit, 4590 tests give "Passed", 419 give "FAILED" and 13 give "Error". On 64 bit, 3669 tests give "Passed", 381 give "FAILED" and 9 give "Error". On 64 bit, many tests are not run (and the SBUs drop down to ~ 30 instead of ~100) in the jdk testsuite, because a segmentation error occurs in the "nio" part:

Passed: java/net/B6499348.java
Passed: java/nio/Buffer/AllocateDirectInit.java
#
# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007f31ab708f0b, pid=14814, tid=139849010583296
#
# JRE version: OpenJDK Runtime Environment (7.0_60-b30) (build 1.7.0_60-BLFS-b30)
# Java VM: OpenJDK 64-Bit Server VM (24.60-b09 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# C  [libjava.so+0x17f0b]  Java_java_nio_Bits_copyToShortArray+0x1fb

I do not know if it is something to worry about. But I am almost sure it was not in previous versions.

[Fernando de Oliveira]

Pierre, may be you have already solved everything I will write bellow, but jsut in case:

First, I noticed that you are going to remove the cacerts patch. I think I do not need to recall that the page must be changed now, perhaps a note is enough, to remember the user to install the cacerts as described in the configuration section. And there, "Install or update" and other sentences need to be modified, now that it is certain that they eill not be installed during build.

Second, the tests need to be run in an X session. If the screen turns off, you have many new failures.

Bruce or Ken once needed to watch the whole test, to move the mouse and prevent failures.

What I do is to install xscreensaver and choose "disable screensaver" or "deactivate screen saver" (in Portuguese it is "desativar o protetor de tela" (tela=écran, screen). Then, failures goes down, because the screen is never turned off.

I don't remember my last test run, but I have for OpenJDK-1.7.0.51-2.4.5 (last numbers are IcedTea version):

$ grep "Test results" OpenJDK-1.7.0.51-2.4.5-jtregcheck-2014.01.29-18h12m38s.log
Test results: passed: 232; failed: 31; error: 3
Test results: passed: 1,956; failed: 2
Test results: passed: 232; failed: 31; error: 3
Test results: passed: 1,956; failed: 2

It was always a mystery why it logs twice.

[Pierre Labastie]

Replying to fo:

Pierre, may be you have already solved everything I will write bellow, but jsut in case:

Not at all, thanks for your input.

First, I noticed that you are going to remove the cacerts patch. I think I do not need to recall that the page must be changed now, perhaps a note is enough, to remember the user to install the cacerts as described in the configuration section. And there, "Install or update" and other sentences need to be modified, now that it is certain that they eill not be installed during build.

How could you know what I was about to forget? ;-)

Second, the tests need to be run in an X session. If the screen turns off, you have many new failures.

I use the instructions with Xvfb as in the book. Looking at the figures below, it seems that it is not too bad.

[...]
I don't remember my last test run, but I have for OpenJDK-1.7.0.51-2.4.5 (last numbers are IcedTea version):

$ grep "Test results" OpenJDK-1.7.0.51-2.4.5-jtregcheck-2014.01.29-18h12m38s.log
Test results: passed: 232; failed: 31; error: 3
Test results: passed: 1,956; failed: 2
Test results: passed: 232; failed: 31; error: 3
Test results: passed: 1,956; failed: 2

It was always a mystery why it logs twice.

This I know why: there are three testsuites, one for hotspot, one for jdk, and one for langtools. They are run one after another. After the hotspot and langtools tests, the test results are output. But for some reason they are not ouput after the jdk tests. At the end of all the tests, a summary is printed, so that you have again the test results. Unfortunately, again not for jdk.

Doing the same as you (32 bit):

$ grep "Test results" test.log
Test results: passed: 241; failed: 41; error: 3
Test results: passed: 1,934; failed: 25
Test results: passed: 241; failed: 41; error: 3
Test results: passed: 1,934; failed: 25

So my results are not too different from yours. But the jdk figures are: passed: 2415; failed: 353; error: 10 (on 32 bit).

[Fernando de Oliveira]

Yes, your results seem very good. Thanks.

[Fernando de Oliveira]

Replying to fo:

Pierre, may be you have already solved everything I will write bellow, but jsut in case:

"Install or update" and other sentences need to be modified, now that it is certain that they eill not be installed during build.

Just remembered: I was wrong: "Install or update" is still good. What is not good: sentences following that, about failure of install during build. Sorry about that.

[Pierre Labastie]

Replying to fo:

Replying to fo:

Pierre, may be you have already solved everything I will write bellow, but jsut in case:

"Install or update" and other sentences need to be modified, now that it is certain that they eill not be installed during build.

Just remembered: I was wrong: "Install or update" is still good. What is not good: sentences following that, about failure of install during build. Sorry about that.

Do not be sorry. I am trying to find a good wording, but I wish I were born in an English speaking country...

[Pierre Labastie]

Hopefully the last question about this update: I read on the page that the test results depend on the cacerts file installed. If I remove the cacerts installation, as discussed above, those tests are sure to fail...

So, should I really remove the cacert installation? (waiting for answer(s), I'll see if it is not too hard to update the patch).

[Fernando de Oliveira]

When DJ suggested to remove, I replied that I would prefer to keep, because I felt it had useful information that could help us learning more.

If I could, I would keep it. If it too difficult to make the patch work, I would remove it, and, if it is easy enough, exclude the respective tests.

I remember it took a while for DJ to include OJDK in the book, there is a lot of work from him in the patches and the page.

Reasons given, I think whatever you decide is good, for me.

[Fernando de Oliveira]

On the other hand, the tests could be run after the package is installed and the cacerts are installed, IIRC.

[Pierre Labastie]

Thanks Fernando. I think I can update the patch. Will delay slightly the commit (I'm so late anyway), but I think you are right in preferring to keep.

Completely different issue. I just read on ​http://blog.fuseyism.com/: This release updates our OpenJDK 7 support with the first release of the 2.5.x series, based on OpenJDK u60. It is the first release to include the PPC64 & AIX port work from SAP & IBM, and also the first not to include the PulseAudio provider, which will now be part of IcedTea-Sound.

Looks like the IcedTea-Sound package has not been released yet (not searched much), but the pulseAudio classes are not in icedtea anymore.

[Fernando de Oliveira]

To be fair: Bruce also did a lot of work in the patches. IIRC, between many things, he modified the scripts Certificate Authorities Certificates and needed to include the modifications in OJDK cacerts patch.

[Fernando de Oliveira]

Replying to pierre.labastie:

Looks like the IcedTea-Sound package has not been released yet (not searched much), but the pulseAudio classes are not in icedtea anymore.

I could find this:

Announcing the Birth of IcedTea-Sound

​http://mail.openjdk.java.net/pipermail/distro-pkg-dev/2014-June/028118.html

Repository (I think it is mercurial, from the hg):

​http://icedtea.classpath.org/hg/icedtea-sound/file/86856b2e3d18

You have the options to download a bz2 or a gz, perhaps remove .hgignore and include in anduin, with or without configure generated (as in fontforge). And create a new page for it.

[Pierre Labastie]

Thanks, will give a try. But, for now, I have trouble with the cacerts file: it is generated during the build, but not copied to the jdk. Since I did not look at that before, I do not know if it is new in 2.5.0. Will try to investigate, but since I can only llok at this during the evening (and testing this build takes sooo long), I am very slow...

[Pierre Labastie]

Well, it IS copied, but I had an error in my script... Fixed at r13322