https://pantsbuild.org/ logo
Join Slack
Powered by
<@UB2J9BQA0> do you recall whether we opened a tic...
# development
w
@hundreds-father-404 do you recall whether we opened a ticket about the fact that the address of singleton/atom file targets like 
python_source
doesn’t end up being a file address? apparently we have an accommodation in place to allow looks with a file address to succeed for those
…which i’ll need to preserve. but want to reference something
no matches for 
address
or 
atom
, although we do have https://github.com/pantsbuild/pants/issues/12917 … but IMO, we should not ever remove it
@hundreds-father-404: do you feel comfortable closing https://github.com/pantsbuild/pants/issues/12917 at this point, in favor of a smaller ticket about this wart?
h
No, no design has been proposed iirc for how to have a 
python_source
target have a file address. Not sure how that would work - you risk a collision with a generated target
apparently we have an accommodation in place to allow looks with a file address to succeed for those
We do?
w
h
at this point, in favor of a smaller ticket about this wart?
I tbh still am generally pro-replacing file addresses with generated target syntax because: 1) consistency, especially w/ adding parametrization to make an even bigger matrix of address variants 2) the 
../../tgt
part But I think I lost this discussion. Most were generally pro file address when last discussed, and you are more fiercely pro-file address than I am anti (and I see your point about copying and pasting, I indeed copy and paste all the time into my IDE to edit tests that are failing. Very convenient)
w
yea, just based on my workflow, it would be a horrible setback not to have a contiguous filename anywhere.
opened https://github.com/pantsbuild/pants/issues/14419 for this… i think that there is probably a good reference for the “remove requiring a name” issue, but i don’t know what it is.
h
altho I'm skeptical my claims "this will work" are true
https://pantsbuild.slack.com/archives/C0D7TNJHL/p1644438087279269?thread_ts=1644437667.231959&amp;cid=C0D7TNJHL Huh...that was definitely meant to be cleaned up iirc. Either way, worth formalizing one way or another. I'm too focused on another thing to remember exactly what the ideal semantics are, but can look if helpful
w
@hundreds-father-404: there are about a dozen tests depending on it, apparently
👀 1
without knowing it, mostly
see e.g.:
Copy code
ValueError: The address `src/protobuf/dir/f.proto` was not generated by the target `src/protobuf/dir:dir`, which only generated these addresses
in https://github.com/pantsbuild/pants/runs/5118546407?check_suite_focus=true
h
hmmm those tests look broken, that's a valid error. lmk if it'd help for me to fix them to unblock you
w
not with this accommodation… i’ve added it back, so they pass
…oh. this was also a thing for 
python_requirement
, huh. i think that that was the case where we last talked about this
h
how so?
w
when you explicitly declare a 
python_requirement
, its name is 
some/explicit:name
rather than 
some/explicit#name
(or w/e)
h
I think I'm confused what "a thing" refers to here? Are you saying we should apply the shortcut to 
python_requirement
for it to be 
#
rather than 
:
?
w
eh, re: `python_requirement`: i might be mis-remembering. i found the thread though: https://pantsbuild.slack.com/archives/C0D7TNJHL/p1637020967476600
👀 1
Open in Slack
PreviousNext
Powered by