← Back to team overview

kicad-developers team mailing list archive

Re: How much do we care about small memory leaks?



FWIW, I've been running the tests with memory checking enabled on both
Linux and Windows (MSVC), and we have several leaks there already, some
from us, some from wxWidgets, some from Gtk, some from Python, and
apparently some from Boost.Test. I don't exactly trust DrMemory yet, it
seems to sometimes show false positives, but I couldn't yet create
simple test cases that have the same false positives.

Since the test cases are low-level stuff with shallow call stacks, those
should give excellent inroads into getting rid of a lot of memory
checker noise.

I've gotten a bit of pushback from the Gtk people on cleaning up their
singletons, apparently they prefer to suppress valgrind messages after
verifying that a particular object is supposed to leak.

What is a bit annoying is that Jenkins seems to be unable to generate
reports for this. We have summary information for our test cases in XML
form, as well as log files from valgrind and DrMemory for individual
cases, but these aren't yet archived or postprocessed.

For report files (but only if no build is currently running :/)



Debug builds on Windows often run into timeouts because memory error
report generation takes close to an hour for some tests. :/


Attachment: signature.asc
Description: OpenPGP digital signature

Follow ups