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
Running make on MacOS yields a bunch of these warnings:
Building executable 'cryptol' for cryptol-2.5.0..
<built-in>:15:10: error:
warning: non-portable path to file '"dist/build/Cryptol/autogen/cabal_macros.h"'; specified path differs in case from file name on disk [-Wnonportable-include-path]
#include "dist/build/cryptol/autogen/cabal_macros.h"
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"dist/build/Cryptol/autogen/cabal_macros.h"
1 warning generated.
The problem is that cabal tries to create the directory dist/build/Cryptol for the compiled modules, and also the directory dist/build/cryptol for binaries. On MacOS, dist/build/cryptol maps onto the existing dist/build/Cryptol, because they only differ by case. I'm using cabal version 2.2.0.0.
I believe that migrating to cabal new-build would get rid of this problem. The current build is somewhat tightly tied to sandboxes, so I'm reluctant to change this in time for 2.6.0, but after that I think it would be very worthwhile to start using new-build in the Makefile.
We're already successfully using new-build on Travis, but we're not using Travis to build distribution tarballs. I think the main remaining step for using new-build for distributions is locating the generated executables. They're deep in the dist-newstyle directory, and I don't think Cabal makes any promises that their locations won't change from release to release.
Running
make
on MacOS yields a bunch of these warnings:The problem is that
cabal
tries to create the directorydist/build/Cryptol
for the compiled modules, and also the directorydist/build/cryptol
for binaries. On MacOS,dist/build/cryptol
maps onto the existingdist/build/Cryptol
, because they only differ by case. I'm usingcabal
version 2.2.0.0.Apparently this is a known issue with cabal: haskell/cabal#4739
Perhaps we should adopt one of the workarounds listed in that thread.
The text was updated successfully, but these errors were encountered: