and Should

Question

< witty crayon 22786> and < ancient vegetable 10556> Should we just punt on directories in While we figure out the cruft we can start removing `immutable input digests` I think

bitter-ability-32190 · Answer

Or do as Stu suggested and just set it to somethign redicuously high

ancient-vegetable-10556 · Answer

annoyingly I am only interested in this ticket because of directories and `immutable input digests` is the solution if directories aren t solvable

bitter-ability-32190 · Answer

Well I d hate to rush things or halt incremental progress for a specific use case

ancient-vegetable-10556 · Answer

If you ve got valuable use case for just files then sure

ancient-vegetable-10556 · Answer

I just don t have a horse in that race

bitter-ability-32190 · Answer

System libraries slightly smiling face

witty-crayon-22786 · Answer

yes i think punting to a followup PR makes sense not because i don t think that we should do it but just so that it doesn t hold up the clearly less troublesome aspect

witty-crayon-22786 · Answer

gt Or do as Stu suggested and just set it to somethign redicuously high or this yea

bitter-ability-32190 · Answer

I guess I worry that if we set it to some high number we just feel good and move on Instead of fixing the roots of the issue

bitter-ability-32190 · Answer

Well lol the native engine is larger than our file limit man facepalming

ancient-vegetable-10556 · Answer

I m shocked wink

bitter-ability-32190 · Answer

I think removing dirs is still good But damn the issue still haunts us Pants dogfooding is full of surprises

bitter-ability-32190 · Answer

In this case I think I ll just re materialize the file not as a symlink in the test

bitter-ability-32190 · Answer

Perhaps we could ve done that for dirs but meh

witty-crayon-22786 · Answer

gt I guess I worry that if we set it to some high number we just feel good and move on Instead of fixing the roots of the issue as long as it actually triggers for `node modules` we can revisit this is a heuristic

witty-crayon-22786 · Answer

but believe me Chris will want to pursue getting it working soon if it doesn t

bitter-ability-32190 · Answer

Any thought on the path huerstic for excludelisting

bitter-ability-32190 · Answer

namely the loose thought of <https github com pantsbuild pants pull 17520 issuecomment 1323965450>

bitter-ability-32190 · Answer

Again my long term worry for the dirs is we really shouldn t be symlinking true source trees as that will likely not get re used So it s worse for people

witty-crayon-22786 · Answer

i would be fine with a configurable list of pathnames as heuristic

bitter-ability-32190 · Answer

< ancient vegetable 10556> for your `node modules` case are those materialized under a source root

bitter-ability-32190 · Answer

Well at any rate <https github com pantsbuild pants pull 17520> now just does files I m hoping for green CI

bitter-ability-32190 · Answer

<https github com thejcannon pants tree bigfile and dirs> is a branch made from the OG before I removed dirs So we can use that to bring back directories