I noticed a regression bug introduced in 8ce1872cd.
I filed a PR that fixes the problem (and explains it) in pypa/pep517#80.
IMO this is a blocker until pep517 will merge/fix it in other way, release a new version and we will update the vendored library.
Lemme know what you think @pradyunsg :smiley:
Nice catch @NoahGorny! ^>^
I think we'd only hit this when pip is invoked as python src/pip ...
. That's a regression that's worth fixing, but doesn't sound like a blocker to me. If it is reproducible using python -m pip ...
or pip ...
, then it's definitely a blocker. It would be great if you could check whether that's the case. :)
I'll go ahead and add this to the milestone, but I won't be marking this as a blocker unless we know that it's possible to hit this failure w/ a not-src/pip
invocation.
There's a whole bunch of related discussion starting: https://github.com/pypa/pep517/pull/55#issuecomment-606791307.
I see, I'll try and test it in the following day!
Nice catch @NoahGorny! ^>^
I think we'd only hit this when pip is invoked as
python src/pip ...
. That's a regression that's worth fixing, but doesn't sound like a blocker to me. If it is reproducible usingpython -m pip ...
orpip ...
, then it's definitely a blocker. It would be great if you could check whether that's the case. :)I'll go ahead and add this to the milestone, but I won't be marking this as a blocker unless we know that it's possible to hit this failure w/ a not-
src/pip
invocation.
I could not manage to reproduce this bug when invoking pip
or by using python -m pip
.
It also seems like it only happens when it is invoked with python3.8 src/pip
and not with python2/python3.6
This is nice, as this means its not a must for this release :smile:
I really think we should drop the suggestion of using python src/pip
from our docs, and suggest developers to use an editable install instead.
Nothing runs pip in that manner, except for developers of pip, during development as per that suggestion. :/
Dropping from pip 20.1 milestone since this isn't a blocker in any way. :)
Most helpful comment
I really think we should drop the suggestion of using
python src/pip
from our docs, and suggest developers to use an editable install instead.Nothing runs pip in that manner, except for developers of pip, during development as per that suggestion. :/