Re: [Tinycc-devel] A little hacky CI for TinyCC

[Christian Jullien]

I LOVE that!!!

If we all agree, is it possible to send an email once and only once with link 
to failure(s) when a new commit is done which breaks at least one supported 
architecture?


-----Original Message-----
From: Tinycc-devel [mailto:tinycc-devel-bounces+eligis=address@hidden] On 
Behalf Of Giovanni Mascellani
Sent: Friday, December 13, 2019 11:54
To: address@hidden
Subject: [Tinycc-devel] A little hacky CI for TinyCC

Hi,

I tried to set up something vaguely resembling a Continuous Integration for 
TinyCC. The only other CI for TinyCC I managed to find seems to be [1], which 
is months out of date and only tests one architecture (amd64).

 [1] https://travis-ci.org/TinyCC/tinycc/branches

My CI tests many architectures. I don't have hardware for all of them, so 
instead I use the runners provided by GitLab CI (which is again amd64), 
spawning a QEMU virtual machine for the tested architecture.
QEMU is not identical to actual hardware, but I believe this is better than 
nothing. Also, the end result is very very hacky and much slower than it could 
be, but, again, the alternative is nothing (correct me if I am wrong).

I currently test for Debian i386, amd64, armhf and arm64. I will add
riscv64 as soon as the latest Debian kernel boots again. In line of principle 
armel too could be added, but I don't know how to generate a Debian armel image 
that boots with QEMU. It would be nice to add Windows and macOS as well, but I 
don't know how to automatically perform tests once the virtual machine has 
started.

Results are here:

  https://gitlab.com/giomasce/tinycc/pipelines

There is a cron job that automatically pushes there commits on the mob branch. 
Incidentally, i386 tests currently fail[2] (I haven't investigated this yet).

 [2] https://gitlab.com/giomasce/tinycc/-/jobs/378533140#L563

For horror lovers among you, scripts powering the CI are here:

  https://gitlab.com/giomasce/tinycc-test

I hope you find this useful for detecting regressions.

Have fun, Giovanni.
--
Giovanni Mascellani <address@hidden> Postdoc researcher - Université Libre de 
Bruxelles