I was digging into the new subsequenttests in calc (thanks Stephan),
and wondering why they are -so- infernally slow; during the slowness we
get 2% CPU utilisation - where I'd expect a 50/50 split Java &
soffice.bin - actually the 46/44% split (I would suggest) points to IPC
being the majority cost of the tests but I digress.
Anyhow - during the slowest bit I get (from)
tail -f core/workdir/unxlngi6.pro/JunitTest/sc_unoapi/done.log
LOG> Trying to select child with index 333
LOG> selected child count: 334
LOG> valid child - true
LOG> Trying to select child with index 334
LOG> selected child count: 335
LOG> valid child - true
...
And ~two of these each second. Digging into the strace -ttt I get
blocks of these calls like this:
488 1322668708.448523 recv(18, "\360\3\26\0\33\0\0003\0\377\377\0\0\1
\362", 15, 0) = 15
... idling around wasting time noisily ...
488 1322668708.954310 recv(18, "\1", 1, 0) = 1
Which looks like the same 500ms sleep we see in the log.
A quick git grep for 'sleep' suggests that this is not happening inside
sc/ (at least not with System.sleep() - unless this is ViewForwarder).
It looks like we're using:
qadevOOo/tests/java/ifc/accessibility/_XAccessibleSelection.java:
log.print("Trying to select child with index " + i + ": ");
The:
private void shortWait() {
try {
Thread.sleep(500);
} catch (InterruptedException ex) {
}
}
Looks like a good culprit in that file.
HTH,
Michael
--
michael.me...@suse.com <><, Pseudo Engineer, itinerant idiot
_______________________________________________
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
