Plain text Report abuse
<kugel> what's up with qnap-jdgordon? * bertrik finally finished reaper man <kugel> how can it do 3 builds taking >100s each while the round was 222s ? <bertrik> Started now with Equal rites <bluebrother> multicore box? <kugel> do you mean multiple clients? <bluebrother> no, I'm wondering if the build server hands out as many builds in parallel as the box has cores. <kugel> IIRC it doesn't <bluebrother> thought that's only guessing -- no idea how the buildserver handles multicore machines <kugel> also, IIUC this qnap is a NAS, unlikely that it has more than 1 core <bluebrother> a box being NAS doesn't tell anything about the number of cores. What makes you believe it's a NAS box? <kugel> they talked about it a while ago (petur and JdGordon ) <bertrik> qnap makes NASs <bluebrother> well, JdGordon might shed light on the issue ... <kugel> the home NASes all don't have multiple cores (looking at qnap's site) <bluebrother> glasschale-Rondom also seems to run multiple builds at once. * gevaerts suspects that something went wrong with either the database log or with the parsing <kugel> gevaerts: the client stats table clearly shows JdGordon's 3 >100s builds. I'd say it's impossible to do those within 222s and without parallel building <kugel> but I'm having paralell builds also sometimes, so far only when a new round starts just after the previous <kugel> (e.g. due to rapid committing) <gevaerts> yes, looks like a bug in the build master <bluebrother> it's also interesting that the graph shows 1s upload time for the bootloaders though bootloaders aren't uploaded. <gevaerts> logs are uploaded <gevaerts> not sure if that's meant to be counted though <bluebrother> hmm, good point. But then I'd expect the upload speed to show something different to 0 in the build round log <gevaerts> ok, clearly a bug in the build master. Those parallel builds seem to get two build commands <gevaerts> and as far as I know, clients shouldn't get parallel builds at al <kugel> I remember that too <gevaerts> (the client supports them, but the current allocation doesn't make use of that) <bertrik> but we do build and upload in parallel, yesno? <gevaerts> yes <gevaerts> see e.g. rb.hostname.be-gevaerts in the latest round <kugel> looks really bad <kugel> it shows that builds are started before being queued (?) <gevaerts> where? <kugel> iuno-bluebrother for example <bertrik> I got some "Connection refused" when connecting to svn.rockbox.org during that round <kugel> either that, or I didn't understand the differences between queued and completed/cancelled <gevaerts> kugel: I think all queueing is done at the start of the round. They seem to be started in the wrong order, but that's different <kugel> gevaerts: is queuing what the build master expects the client to do (best-fit) <kugel> ? <gevaerts> kugel: yes, more or less <kugel> ah that makes sense <kugel> then iuno-bluebrother totally broke <kugel> wrong order and parallel builds (maybe that's connected?) <gevaerts> I'm pretty sure the wrong order is a symptom of the parallel builds problem * Grahack hat die Verbindung getrennt ("Leaving.") <kugel> gevaerts: I'm not sure it's a bug in the master, it seems to allocate correctly * Topy44 hat die Verbindung getrennt () <gevaerts> I'm not sure. Let Zagor find out... * kugel likes the Wiktionary entry for "automagically" <bluebrother> well, that was part of one of my quesition: does the time for queued builds name the time that the client is supposed to need for the build?