| Newsgroups |
gmane.linux.lfs.beyond.support |
| Message-ID |
<[email protected]> |
On Sun, 2025-08-17 at 21:20 -0500, Bruce Dubbs wrote:
> On 8/17/25 8:38 PM, Randy McMurchy ([email protected] via blfs-
> support Mailing
> List) wrote:
> > Hi all,
> >
> > I had an issue when building the WebKetGTK package (linked to GTK-3) and perhaps this
> > is worth mentioning. I'm building using a Dell Latitude machine that has an 8th
> > generation Intel 8665U Whiskey Lake processor, 16GB of RAM, and an SSD drive. (yes, I
> > realize it is not quite modern). Note:
>
> [snip]
>
> > The instructions for building the package include a note about machines with less
> > than 16GB of RAM need to pass a cmake option so the build won't fail. Well, my
> > machine failed to build due to what looks like is described in the note. I did not
> > use the cmake option as I have 16GB of physical RAM. After looking at sar stats for
> > CPU and memory, it was obvious to me that about halfway through the build, the
> > machine stopped and required rebooting. I deleted the build directory and restarted
> > the build this time using the option mentioned in the note. It did not help, as it
> > failed again at the same exact spot (about halfway through the build).
>
> [snip]
>
> > The second time to build (after using the option to cmake from the note in the
> > package instructions) I realized it was locked up but I managed to use another pty
> > and killed the build process so I didn't have to restart the machine again). I have
> > the MAKEFLAGS envvar set to -j8. I've never had any problems building anything in LFS
> > or BLFS, including LLVM/clang and QT6. The system has been solid as a rock. I have
> > Ubuntu on the host machine at it ships with a firmware updater and everything on this
> > system is up to date.
> >
> > So I decided to build again (and again using the option to cmake shown in the note),
> > but this time passed -j4 to the ninja command line. The build was successful, though
> > painfully slow. Note:
>
> [snip]
>
> > So, unless the machine I'm using to build just simply has insufficient resources, it
> > might be a good idea to add something to the WebKetGTK instructions that warns there
> > are times using some machines that building with 8 parallel jobs can cause the build
> > to fail. I did not try to research the -dev archives to see if this situation has
> > already been identified and discussed. If it has, my apologies for the noise. I'm now
> > going to build the package again, this time linking GTK4 libraries, without using the
> > cmake option but I will pass -j4 to ninja. I will update when I'm finished.
>
> I removed a lot of the data to make this a little shorter. First of all, webkit is
> one of our most intensive packages. There are 12 packages in BLFS with times that
> exceed 10 SBU. At 21 SBU webkit is the third most time consuming packages. Only
> libreoffice at 24 SBU and qtwebengine at 45 SBU are larger. High SBU times indicate
> lots of RAM usage also. Running at j8 on a 4 CPU system will make things worse.
The CPU has SMT so there are 8 logical cores. Thus using -j8 is proper
unless SMT is disabled via the BIOS or the kernel cmdline.
> Two questions. How much swap do you have? Did you check dmesg for OOM (out of
> memory) conditions terminating some programs?
The problem is using swap can cause the build to be significantly
slower, likely even slower than reducing the number of parallel jobs.
I'll reword the <important> box a little.
--
Xi Ruoyao <[email protected]>
--
http://lists.linuxfromscratch.org/sympa/info/blfs-support
Unsubscribe: See the above information page