On Mar 22, 2018 22:52, "Kieran Duggan" kieranduggan15@gmail.com wrote:
Would making the tests safe to run in parallel be a good proposal idea? Or is it too ambitious for a summer project. Running these tests in one process is painstaking on my FX-8350
On Thu, Mar 22, 2018 at 11:47 PM, Kieran Duggan kieranduggan15@gmail.com wrote:
That makes sense
On Thu, Mar 22, 2018, 11:45 PM Austin English austinenglish@gmail.com wrote:
On Mar 22, 2018 22:42, "Kieran Duggan" kieranduggan15@gmail.com wrote:
Cool I didn't get the chance to take a look at it before. I'm using it now. I noticed that the line: time make -k test >> "$logfile" 2>&1 || true doesn't use the -j option for make. I'm not sure if this is intentional, but when I changed it to time make -j"$(nproc)" -k test >> "$logfile" 2>&1 || true it ran much faster. Thoughts? Am I just breaking something without realizing it?
Many tests aren't safe to run in parallel, things may fail.
Imo probably not a GSOC project, but I'm open to disagreement.
The full tests take about 20 minutes, but valgrind takes 10 hours (on my machine). I wouldn't recommend that to check on a single bug. Using the wrapper and suppression files should get you a setup close to mine for testing, however.