Lombix OS / Blog: Recent posts

After a quick build of Lombix 2019 9 ( which I have now running on my workstation ) I now will atempt to build 2020. And again I will try to compile everything in their own directory. I find this very very useful. Also, I need to get into the "containers" universe so I am including now go, runc and containerd by default.

I never really bothered to have sound within QEMU. But just today I wanted to test something in the new image , and I finally decided to look for the switches necessary to do so. It is actually trivial, with some caveats. This is what I run

qemu --enable-kvm -m 4096 -soundhw hda lombix-2019.img

That emulates the HDA codec that I have in my desktop, and with that I can try to sort out some minor issues with the snd-hda-intel module.

I want to explore the possibility to use cairo-dock cairo-dockinstead a taskbar , so I compiled cairo-dock and the cairo-dock-plugins. Problem was that cairo-dock-plugins was failing because some errors with the alsamixer plugin. At the beginning I was a bit confused because I didn't know why ( I do not like cmake , I still do not understand how it works ) , but the truth was yelling at my face, the error was a redefinition of the structs timeval and timespec.
Long story short, I read about and I've found this case in stackoverflow in stackoveflow that gave the hint for a workaround and be able to at least compile successful the cairo-dock-plugins... read more

Well, I've been using Lombix 2018 at home for a while, and I think it is time to prepare for 2019, so I will abandon 2018 ideas and start the new Lombix 2019 with a new approach. First, I will abandon the "every package in their home directory" , it is too costly to maintain. I will try an "intermediate" approach, /usr should contain the "basic core" up to the Xorg , and then in /opt splitted in several folders some a few areas. I will explore this idea, yes.
Also I am going to "compile" more packages. At this moment I have an image with almost everything ready, but firefox, I've hit again the annoying issue of the "fatal error: cstddef" which I've faced several times but never bother to write down what is the cause. I suspect it is something to do with CPLUS_INCLUDE_PATH environment variable. When I fixed it this time I will write a few lines here to remember

Uff, it took me some effort, but finally I managed to compile Libreoffice 6. I was having an strange annoyng problem when using gpgme, the C compiler could not find the iosfwd C++ header, and I never understood why the C wanted something from the C++. The problem was with the internal xmlsec package. So I decided to find the XMLSEC myself, and compile it apart, and tell Libreoffice to use that library instead its internal copy. BTW: I had to compile pk-11 as well, and I did. After that I tried again and it was done, just for a few bits, as it took more than 24 hours within QEMU to compile and it got almost 100% of the 16Gb partition disk space.
A warning though, I have not tested it properly, so I would not be surprised if it crashes under normal use, but hey, the most difficult part has been done.
With this I am a bit closer to move the status of the project to BETA... but not yet.

I've moved the status of the project from preAlpha to Alpha. I think the image is of enough quality to be considered it such. At this moment I am trying to install openblas, numpy scipy and finally the scikit. That's the kind of stuff I am interested these days. However, I actually cannot move towards beta stage until I solve the annoying problem of the Libreoffice, which refuses to compile.

It took me a while. The problems were due to "corrupt" system headers. I cleaned them up, also I use the chance to upgrade to the latest kernel 4.15.14 . I do still have an issue with ALSA modules , I need to unload the snd_hda_intel and load it again with modprobe. Before using the modules feature on boot I would like to understand why this is happening. I suspect some issue with how I compile the kernel ( see the config file in /usr/boot ).... read more

I am puzzled now with ffmpeg and mplayer. They fail to compile when I was pretty sure they did in the past. It seems the problem is that within Qemu both packages seem convinced that SSSE3 cpu flags are available but a quick look at /proc/cpuinfo it is obvious they are not. At some point I used the qemu parameters --cpu host to check a few things, this means I compiled some packages using my native CPU ( the ssse3 was available ). It should not a big problem, I can recompile all the stuff again using the Qemu default cpu, but , no. Now, ffmpeg INSISTS that it has detected SSSE3 instructions and, of course, it crashes.... read more

I am now, that I've gave up with LibreOffice, fixing all sort of bugs/issues in Lombix. This entry is just to track one specially anooying, the lack of sound in Firefox 58 when compiled to use only ALSA. It is true this is a "no future" path , Firefox has dropped support to ALSA since 52 ( or 53, not sure ) , but the code is still there and if you use the compilation instructions from BLFS you can get it. In my case the instructions from BLFS are not complete, in BLFS they say that you need to change one parameter in Firefox ( using the about:config page ). But I've found that I actually have to change three:
~~~
security.sandbox.content.level 0
security.sandbox.content.syscall_whitelist 16
security.sandbox.content.write_path_whitelist /dev/snd/
~~~... read more

I am still working in Lombix, at this moment I am stuck with LibreOffice. The compilation crashes because it cannot find the iosfwd header file. I do not understand why it such thing happen but eventually I will get the solution. Problem is that every attempt it takes a whole day, libreoffice is perhaps the heaviest package to compile so far.

Also, I want to upgrade Firefox to at least 58 version. People I know told me that the performance improvement is more than evident, so let's try. ... read more

Firefox 55.0.3 has been finally compiled. the problem was related to the final step where the make install wanted to do strip on the binaries. For some reason I don't understand this was not working. I suspect the make install was unable to find the binutils strip command and somehow replaced this by "yes" command and the it got into a never evending loop. The solution was apply disable-strip to the configuration , as the lmbx installation tool already does an strip.
Now, I am having similar issues with qemu, it is one of the final packages that remain, together with libreoffice and php. Once I have these three under control I will move the release status to ALPHA. ( I am labelling the current work as pre-ALPHA) .
So, perhaps 1 week more, 2 weeks, I cannot say. I will have the first ALPHA release.... read more

Well, that's kind of a pretentious title, but what a relief. Cmake has been a nightmare for me as I wasn't understanding how it works and because of that I was doing some very ugly hacks and patches. I started to look at the source code and then I've found the function find_patch, and , you know what? It ocurred to me RTFM !!!. Yes, I search for "cmake find_patch" and I got the cmake documentation https://cmake.org/cmake/help/v3.0/command/find_path.html where the solution is simple. Use CMAKE_PREFIX_PATH.
I've recompiled CMAKE without nasty hacks, I've recompiled openjpeg, jasper, and I finally was able to compile x265 which was one of the packages I wasn't able to build. Now everything works. Finally.

I am writing here a quick note about an issue I was having compiling cyrus-sasl. I was getting a compilation error because the make said it didn't know how to compile "-lgdbm". I have to clarify that I am not using BerkelyDB, just gdbm instead.
It seems the problem was in the LFS patch cyrus-sasl-2.1.26-fixes-3.patch , around the line 743,

~~~
-libsasldb_la_DEPENDENCIES = $(SASL_DB_BACKEND)
-libsasldb_la_LIBADD = $(SASL_DB_BACKEND)
+libsasldb_la_DEPENDENCIES = $(SASL_DB_BACKEND) $(SASL_DB_LIB)
+libsasldb_la_LIBADD = $(SASL_DB_BACKEND) $(SASL_DB_LIB)
~~~... read more

At this moment I am trying to compile Firefox 55, it seems ( only it seems ) that might be compiled without pulse audio and work only with ALSA. The instructions of LFS says PulseAudio is required by it still can be edited the mozconfig , and that's what I am doing. It can compile, right, but a different matter is any sound is generated. I will like to see that by myself first.
However, Firefox is one of the most heavy pieces to compile, you have Boost, Rustc, LibreOffice and Firefox, as the 4 big ones. I've already "pre-compiled" rustc because it is a must for Firefox, the other two will be attemted later.
Once Firefox is compiled and a few validations are done I will release it as the next version, I think the first Release Candidate is taking form and I would like to include Boost and LibreOffice.. and that's it , RC1 is out. Perhaps one or two weeks more of work.

I've wasted two days reconsidering the layout of Lombix. Finally, I've thrown away all that work and resume where I was before, well, except some minor changes I am applying now. I will further the "every package in their own folder" also splitting man , doc, pkgconfig files, and others away from share and into each individual package. This means making use of env variables like ACLOCAL_PATH PKG_CONFIG_PATH. I will be recompiling all packages again to relocate things then way I want. Of course, other "shared" folders like icons, dbus-1 , glib will need some tricks ( aka symbolic links )

Firefox now requires the new language/platform rustc, I have all other requirements ready ( I will try to avoid pulse audio using the mozconfig file, let's see if that works ) , but rustc is probing a difficult challenge. Because to boostrap rustc you need to download some binaries files you can imagine these files were compiled "assuming" the classical Linux setup. So, "/lib64" was required, the SSL_CERT_FILE was also required to help him found the CA Certificates, and the msot dirty trick, I had to create a link in /usr/pkgs/gcc/bin for "ar" , as it was insisting to find the ar command in /usr/pkgs/gcc/bin
At this moment is "compiling" , don't know yet if it will finish succesfully , but it looks like a monster.

I can see someone saying that autotools might not be as friendly as they could be, and also I can see perfectly as someone trying to do something about, but, cmake ? When I compile/install libtool+autoconf+automake is rather trivial, simple and can understand a few things. When I compile cmake is a big monster that doesn't want to be tamed. When I compile something that was prepared with the autotools , I know configure is my friend and if something doesn't fit then LDFLAGS, CFLAGS and so on. With Cmake that doesn't work. I know , I know, it is a matter of knowing the tool and learning new tricks.... but, we already had something that worked.
Yep, I am preparing cmake for Lombix.

Well, there is still a lot of do, but I am at a stage where I have to start to install GTKx and , perhaps, QT. It annoys me the situation of GTK, specially that decission of the Client Side Decorations ( CSD ) . I am not sure about GObject Introspection, perhaps there were good technical reasons to do that, but forcing CSD, just like that, without being an option, was not a particularly nice decision.... read more

I read the other day that 4.15 was out, and apart from the well known issues of the Spectre and Meltdown vulnerabilities I saw in the release notes this new fancy cool feature

A /sys/devices/system/cpu/vulnerabilities/ directory has been added that will show the vulnerabilities affecting your CPU and the mitigations being currently applied.

So, I've compiled 4.15.1 in Lombix, runnig under Qemu, as I understood that Meltdown was addressed in this patch thanks to KPTI and the Spectre V2 was also address with the retpoline, but, when I search myself into this directory I got... read more

I've learned that SourceForce can actually host your site, apart from ths Wiki and space for files I: can host an entire site. So, today I've registered the doamin http://lombix.org and I've redirected it to a Vhost here. At this moment I am not doing more thant that, but soon I will be investigating that functionality and setup a database ( pity, it has to be mysql, I would rather go for a postgres ) and drupal, all hosted ( if I understood correctly ) by SourceForge with no evident restricitions in terms of disk space.
In the first image , right now I am recompiling everything again to place the things the way I wanted, I: guess I would need 1 month more before I published the first image ( raw or qcow, don't know yet )

It has been a while since I setup this project in sourceforge. I've been debating myself since then whether it was a good idea to make Lombix pubic or not. I finally decided to do so, perhaps I might some ideas from someone, working alone sometimes is a bit boring.
In the next days I will release the first qemu image , a very limited version of Lombix just to have "something" to show, then we will continue from there

Hi, I've just set-up the Sourceforge account and I am still in the process to fill in all the forms required, and in the process learn about this, so this is not a very relevant entry, indeed.