Lukas Straub <lukasstra...@web.de> wrote: > On Thu, 20 Apr 2023 12:20:25 +0200 > Juan Quintela <quint...@redhat.com> wrote: > >> Lukas Straub <lukasstra...@web.de> wrote: >> > Add postcopy tests with compress enabled to ensure nothing breaks >> > with the refactoring in the next commits. >> > >> > preempt+compress is blocked, so no test needed for that case. >> > >> > Signed-off-by: Lukas Straub <lukasstra...@web.de> >> >> Reviewed-by: Juan Quintela <quint...@redhat.com> >> >> And I wanted to removed the old compression code and it gets new users. >> Sniff. > > Who know how many compress threads users are out there...
Not too much. We broke it during development and nobody found it. And the reason that I wrote the multifd-zlib compression code was because I was not able to get a migration-test working with compression, so .... > By the way, I'm not against deprecating compress threads in the long > run. I'm already working on (cleanly :)) adding colo support with > multifd. Ok, then I will still put the deprecate comment there. >> > --- >> > tests/qtest/migration-test.c | 83 +++++++++++++++++++++++------------- >> > 1 file changed, 53 insertions(+), 30 deletions(-) >> > >> > diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c >> > index 1f2a019ce0..930cb4f29d 100644 >> > --- a/tests/qtest/migration-test.c >> > +++ b/tests/qtest/migration-test.c >> > @@ -1127,6 +1127,36 @@ test_migrate_tls_x509_finish(QTestState *from, >> > #endif /* CONFIG_TASN1 */ >> > #endif /* CONFIG_GNUTLS */ >> > >> > +static void * >> > +test_migrate_compress_start(QTestState *from, >> > + QTestState *to) >> > +{ >> > + migrate_set_parameter_int(from, "compress-level", 1); >> > + migrate_set_parameter_int(from, "compress-threads", 4); >> > + migrate_set_parameter_bool(from, "compress-wait-thread", true); >> > + migrate_set_parameter_int(to, "decompress-threads", 4); >> > + >> > + migrate_set_capability(from, "compress", true); >> > + migrate_set_capability(to, "compress", true); >> > + >> > + return NULL; >> > +} >> >> Independently of this patch, we need to change this test to use 4 >> compression tests and 3 decompression or anything that is not the same >> number in both sides. >> >> I was complaining about this and when I arrived to the end of the path >> found that this was code movement. >> >> Later, Juan. >> > > Oops, forgot to mention, the test is based on this patch > https://lore.kernel.org/qemu-devel/2f4abb67cf5f3e1591b2666676462a93bdd20bbc.1680618040.git.lukasstra...@web.de/ > > Will probably carry the patch with this series then. So you mean 4 > compress _threads_ and 3 decompress _threads_? Yeap. Later, Juan.