On Sat, 27 Jan 2024 12:08:59 +0200 Kalle Valo wrote: > Jakub Kicinski <kuba@xxxxxxxxxx> writes: > >> I don't run checkpatch except for ath10k/ath11k/ath12k, too much noise. > >> I ended up adding this to my script: > > > > We run build with sparse and W=1 and then diff the number of warnings > > to weed out the pre-existing ones, FWIW. > > So for wireless and wireless-next I now check W=1 warnings every time I > push. We are mostly warning free now but I'm not checking the linker > warnings, for example the current MODULE_DESCRIPTION() warnings. > > It's really annoying, and extra work, that people enable new W=1 > warnings before fixing them. Could we somehow push back on those and > require that warnings are fixed before enabling with W=1 level? My quite possibly incorrect understanding is that "giving people time to fix" is the main point of W=1 :( W=2 is for stuff which may false positive, W=1 is for stuff which does not false positive but we can't enable it in formal builds because the tree is full of it. > In wireless there is a significant number of sparse warnings. I have > tried the cleanup people to fix them but it seems there's no interest, > instead we get to receive pointless cleanups wasting our time. <loud sigh> Tell me about it.. :) > BTW the 'no new line at end of file' warning is indeed from sparse, like > Arend suspected: > > drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwil.c:432:49: warning: no newline at end of file BTW I'd happy to help you set up an instance of our build testing bot, if you have a VM that can be used. It does take a bit of care and feeding, but seeing the build failures in patchwork pays the time back.
