On Wed, 9 Mar 2022, Eric Pouech wrote: [...]
sorry, my post wasn't clear... I meant replacing the serie of 25 patches with:
one serie of 4 patches (user)
21 individual mail for each of the changes (not folding all of those in a
single mail)
That will help the debian11 VM a bit.
The main bottleneck is the debian11 VM on VM1, though w1064 on VM3 is not far behind.
The statistics page says that the (non-WineTest) debian11 tasks take 5m 11s on average. Add the 43.6 second revert time and that means it can process a bit over 10 tasks per hour.
Each email generates 2 debian11 tasks (one for 32-bit and one for wow32+wow64). So 25 emails will generate 50 tasks and thus take about 5 hours to process.
Another tidbit: over the past 7 days debian11 has received 5.3 tasks/hour on average. That's 127 tasks/day, though my gut feeling is that most of these have been during weekdays so it's more a rate of about 170 tasks/day (7.4 tasks/hour).
if there are niceness issues, another option is to send less patches every day
(I tried to keep it ~25 / per day, even it's more ~20 lately)
It's not so much the number of patches per day as the number of test units patched.
so reducing to one mail = one patch = one file (in tests/*.c) will help here?
Yes.
Looking at the stats page I noticed that the average debian11 revert time is a bit long (43.6s) so I'm doing a refresh of the VM. That will reduce the size of the disk image and should speed up the reverts a bit (I'm hoping for reverts in the 10 to 30s range).
At the same time I'm upgrading the VM from Debian 11.0 to 11.2.
And I'm also trying to update debiant: there's little point to a Debian Testing machine if it's completely out of date. But at the same time OpenGL multiarch support is quite broken in Debian Testing so the latest may not be very usable. Good to know either way.