<@U06A03HV1> <@UB2J9BQA0> trying to run `build-sup...
# development
p
@witty-crayon-22786 @hundreds-father-404 trying to run
build-support/bin/generate_lockfile.sh
for https://github.com/pantsbuild/pants/pull/10803 and the pants hangs... here is the output (I modified the script to output debug messages):
w
looking
are you using python3.8?
p
seems that pants is using 3.8.
w
try passing
PANTS_PANTSD=false
p
Screen Shot 2020-09-17 at 3.15.37 PM.png
w
or exporting it in the relevant shell
p
that worked...
👍 1
w
yea. 3.8 technically isn’t supported yet. i’m going to reopen the ticket we had about this. greg wasn’t able to repro, but that could be because he’s on linux.
👍 1
p
running
./pants --tag=-lockfile_ignore dependencies --type=3rdparty ::
worked... but it probably returned results from cache, so it probably doesn't count...
let me try on a clean branch (with a different reqs.txt file)
w
the issue is that it is piped somewhere, i think.
p
yeah... probably...
h
benjy also had trouble repro-ing on his mac
I happen to be looking at https://github.com/pantsbuild/pants/blob/master/src/python/pants/pantsd/pants_daemon.py#L38 in the process of trying to clean up pantsd logging initialization - I wonder if this could be related
👍 1
👀 1
I haven't dug much yet into how this class works, but it involves messing with
sys.stdout
, so it seems possible?
w
i’m able to repro. will take a look.
👍 1
mmm, i was wrong about which issue it was… this is https://github.com/pantsbuild/pants/issues/5510
much older, and already triaged as a client hang… so it’s likely that porting to the rust client will resolve it.
🦀 1
👍 1
also had occasion during debugging to try out the pure rust client again:
Copy code
$ time nails-client 127.0.0.1:56164 ./pants list :: | wc -l
     254

real	0m0.266s
user	0m0.007s
sys	0m0.009s
💯 1
one of these days. sigh.
@hundreds-breakfast-49010: were you still thinking of doing that port to the nails client library?
h
yeah, basically contingent on whenever it makes sense for me to spend the time to do so
do we know why this problem only manifests itself on python 3.8 on os x,and only sometimes?
w
see my comment on the issue: it repros on 3.7 with the alternate repro
the alternate repro is what we already had a ticket about