<somlo>
dxld: mine failed the same way on both f38 and rawhide (f40), I'll try re-running the mock builds again tomorrow, but don't expect to get a different result...
unkraut has quit [*.net *.split]
agg has quit [*.net *.split]
acathla_ has quit [*.net *.split]
bpye has quit [*.net *.split]
unkraut has joined #yosys
acathla has joined #yosys
bpye has joined #yosys
agg has joined #yosys
<dxld>
all my build problems, first the test then some latex errors, have mysteriously dissapeared now. gah :/
citypw has joined #yosys
FabM has joined #yosys
FabM has joined #yosys
phung has joined #yosys
phung is now known as kevin998x
kevin998x has quit [Quit: WeeChat 4.0.4]
phung has joined #yosys
phung is now known as kevin998x
<jix>
getting core dumps or at least backtraces for those crashes would be helpful to figure out what's going on
kevin998x has quit [Quit: WeeChat 4.0.4]
<dxld>
well it's not usually crashes, just errors.
<dxld>
now I got a race on the compilation of tests/tools/cmp_tbdata (Text file busy error)
<jix>
well in those cases anything else that would help reproduce it
<jix>
ok for cmp_tbdata I can see how that could result in a race condition and will look into how to fix that
<jix>
for failing tests it can be helpful to get a full log of make test and a tar of the tests directory (or a subdirectory if it's clear that the failing test is fully contained in it)
<jix>
anything abc related can be quite hard to reproduce though, if you happen to have a system (little endian and linux) where you can reliably reproduce abc related failures I'd be interested in figuring out a way for me to reproduce such an environment
<jix>
but it really looks like something the makefile should handle
<dxld>
jix: I prefer flock but perhaps that's not portable.
<jix>
dxld: yeah, doesn't seem to exist on mac os which we do support
<dxld>
I think the -nt (newer than) thing is a dead giveaway this should be in the Makefile :]
<jix>
yeah that's exactly what I was thinking :D
<dxld>
BTW would it be possible to move the list of test invocations to a script? right now I have to munge the test target depends to allow testing yosys "as-installed" without rebuilding everything needlessly
<jix>
the problem is that the execution goes via makefile -> shell script -> makefile -> shell script and this would have to be a target in the outer makefile, but the dependency on that would be in the inner one AFAICT
<dxld>
I mean is anyone going to expect cmp_tbdata to get rebuilt when they run the tests directly?
<jix>
dxld: I have no idea what cmp_tbdata actually does, so maybe not
<jix>
ok looking at it, I think it's fine to just have it built by make test
<dxld>
I would build it as part of the all target actually
<dxld>
that way you can `make all && ./run-tests.sh` without having to worry about compiling this helper
<dxld>
but having the test target depend on it too is fine obv
<jix>
I think it's fine to have it build by make all, and I also wouldn't mind moving the make test commands to a script but I'd keep it within the tests directory
<jix>
(as for test subdirectories the script to run the tests is also contained in that direcotry so it's more consitent that way IMO)
<dxld>
fine with me
<jix>
(and the make test target should remain, but just invoke that script instead of listing the commands in the makefile)
<dxld>
yeah
<dxld>
only problem I see with this setup is paralellization across of test dirs, which I've been meaning to add
<dxld>
(work imbalance sucks for speedup reasons)
<dxld>
hmm but that should be fine actually, just need to pass the make jobserver down which is easy enough
somlo_ has joined #yosys
somlo has quit [Remote host closed the connection]
somlo_ is now known as somlo
<somlo>
jix: I'll re-run the builds, and collect the core file(s) if, as I expect, a similar crash will happen again
daglem has quit [Server closed connection]
daglem has joined #yosys
srk has quit [Server closed connection]
srk has joined #yosys
lumo_e has joined #yosys
lexano has joined #yosys
lumo_e has quit [Quit: Quit]
strobo has quit [Ping timeout: 246 seconds]
strobo has joined #yosys
so-offish has joined #yosys
greeb has quit [Read error: Connection reset by peer]
greeb has joined #yosys
greeb has quit [Client Quit]
greeb has joined #yosys
citypw has quit [Ping timeout: 246 seconds]
FabM has quit [Ping timeout: 258 seconds]
<somlo>
apparently my problem is related to commit 9f8e039a4 ("ast: use new format string helpers.") and something about Fedora's build environment (CFLAGS) -- I submitted issue #3923