I think I agree. The tags are a good short term solution to keep those tests from blocking things, although it would be best to minimize host dependencies in the longer term. That's been problematic but I think unavoidable with KVM resulting in it being perennially broken on x86, so ideally we can shoot for something that works at least good enough to test on any/most hosts.
The danger would be making SE too fancy and too much like a weird version of FS which makes it make less sense as its own way to run gem5, but I think we can probably avoid that if we think things through carefully. But since that might be time consuming and at least a little non-trivial, I think the tags are a reasonable thing to do in the mean time. Gabe On Thu, Jan 16, 2020 at 8:53 AM Jason Lowe-Power <ja...@lowepower.com> wrote: > Hi Giacomo, > > The third option seems good to me. In fact, we could have another tag in > the regressions for the kind of host to run on. Right now we have "quick" > and "long", but we could also add other tags specific to hosts. > > I think it's fine to run different tests on different hosts. > > Cheers, > Jason > > On Thu, Jan 16, 2020 at 5:48 AM Giacomo Travaglini < > giacomo.travagl...@arm.com> wrote: > > > Hi devs, > > > > As you might know, gem5 is running a dynamically linked hello world as a > > regression for x86. > > This works fine when it is run on a x86 host. > > > > However this will prevent regressions to succeed when run on a non-x86 > > host. > > In order to make it work there are several options: > > > > 1) Use --redirects for se.py (This is the dynamic linking support I > > introduced some time ago) > > to point the interpreter to x86 libraries. This is feasible but will > > pollute the code a little bit and it > > is not strongly portable. > > > > 2) Remove hello-dynamic from regressions: Nothing to say, I'd rather stay > > away from this option though > > > > 3) Add an extra dynamic/static TAG: So that when running regressions on a > > non-x86 host the > > --exclude-tags dynamic will be used. This is more portable than 1, but > > will create the case where a > > different number of tests will be run depending on the host machine. > > > > If I have to be honest I would go for the third option; what's your > > opinion about it? > > > > Giacomo > > IMPORTANT NOTICE: The contents of this email and any attachments are > > confidential and may also be privileged. If you are not the intended > > recipient, please notify the sender immediately and do not disclose the > > contents to any other person, use it for any purpose, or store or copy > the > > information in any medium. Thank you. IMPORTANT NOTICE: The contents of > > this email and any attachments are confidential and may also be > privileged. > > If you are not the intended recipient, please notify the sender > immediately > > and do not disclose the contents to any other person, use it for any > > purpose, or store or copy the information in any medium. Thank you. > > _______________________________________________ > > gem5-dev mailing list > > gem5-dev@gem5.org > > http://m5sim.org/mailman/listinfo/gem5-dev > _______________________________________________ > gem5-dev mailing list > gem5-dev@gem5.org > http://m5sim.org/mailman/listinfo/gem5-dev _______________________________________________ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
