[Libwebsockets] Sai administrivia

Andy Green andy at warmcat.com
Mon Nov 2 17:50:33 CET 2020


Hi -

If you don't care about the issues and solutions maintaining 21 OSes for 
libwebsockets CI, you can just ignore this kind of email.

I updated the big AMD machine I bought for CI a few months ago with a 
1TB SSD, and it has a new location where I can't hear the fans so it 
should be up all the time.

There are 18 OSes in there, mainly systemd-nspawn "high-functioning 
chroots" with private subnets but also Qemu instances for freebsd and 
windows.

I also started updating the host OS to Fedora 33, this was fine except 
with the update came a necessary rejigging of Risc-V Qemu to know about 
bootloaders natively, rather than have to concatenate the bootloader 
with the kernel image, breaking all Risc-V images that predate it.

There's a simple workaround on qemu commandline but the qemu instances 
all live in libvirt, which wants all settings in XML... for posterity 
the necessary blob at the end is

at the top

<domain type='qemu' xmlns:qemu='http://libvirt.org/schemas/domain/qemu/1.0'>

then at the end before the closing </domain>

   <qemu:commandline>
     <qemu:arg value='-bios'/>
     <qemu:arg value='none'/>
   </qemu:commandline>

After that I updated the vm to f33 Risc-V arch via rpm and rebooted, it 
works well but there's another f33 problem, the rpmbuild macros have 
been updated to include -ffat-lto-objects, which risc-v doesn't 
understand and the RPM packaging part of the CI dies.

Solution is create /etc/rpm/macros and put

%_lto_cflags %{nil}

in there.

-Andy


More information about the Libwebsockets mailing list