On Thu, Oct 29, 2020 at 02:52:16PM +0000, Peter Maydell wrote: > On Thu, 29 Oct 2020 at 14:31, Christian Schoenebeck > <qemu_...@crudebyte.com> wrote: > > > > On Donnerstag, 29. Oktober 2020 15:15:19 CET Peter Maydell wrote: > > > On Thu, 29 Oct 2020 at 14:06, Christian Schoenebeck > > > > > > <qemu_...@crudebyte.com> wrote: > > > > Ok, I'll use mkdtemp() instead, that avoids other potential parallel > > > > config > > > > colissions that I may not have considered yet. > > > > > > > > The original motivation against mkdtemp() was that /tmp is isually a > > > > ramfs, > > > > hence very limited regarding large file tests. But that's not an issue > > > > right now. > > > > > > How large is "large" here ? > > > E.g. ~10 GiB which might be a problem for cloud based CI platforms. > > Yeah, 10GB is too big by an order of magnitude for anything in the > standard "make check" set. It could go in an optional "I'm the 9p > maintainer and I want to run a wider set of tests" target though.
I think as a rough rule of thumb, tests should not create files that are larger than the size of the QEMU build dir, and if it is creating non-trivially sized files, then they should be created in the build dir, not /tmp. This probably puts 1 GB as a upper bound on size but bear in mind tests can run in parallel, so ideally biggest file sizes should be more in the 100s of MB range Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
