You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
cabal doesn't pass -j to GHC, to avoid launching too many processes. But that's the case even when it builds one package, in particular, the last one, after all dependencies are done. That's annoying when running cabal install on big packages, say Agda.
And while this is clearly a hack, it guarantees that not too many processes are created.
By #976 (comment), cabal build can pass -j but cabal install doesn't.
The text was updated successfully, but these errors were encountered:
It seems to me, if I specify two packages explicitly, that leads to too many jobs. And if I use --only-dependencies, that leads to too few jobs.
The "hack idea" is "ensure trivially that at most one ghc -j is running", where "trivially" means "with limited code changes" and with a trivial proof.
A generalization consistent with this would be to use GHC -j also during the build, if there's a single package in the package queue. That can happen (seldom) before the end when all further packages depend on one. Is this easy/feasible to implement?
cabal doesn't pass
-j
to GHC, to avoid launching too many processes. But that's the case even when it builds one package, in particular, the last one, after all dependencies are done. That's annoying when runningcabal install
on big packages, sayAgda
.And while this is clearly a hack, it guarantees that not too many processes are created.
By #976 (comment),
cabal build
can pass-j
butcabal install
doesn't.The text was updated successfully, but these errors were encountered: