Details
Diff Detail
- Repository
- R11 FreeBSD ports repository
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
| devel/hs-ghcup/Makefile | ||
|---|---|---|
| 12 | Style nit: USES should be sorted. | |
| 185 | Hum, now that we have microlens-0.4.14.0 in USE_CABAL, these constraint flags shouldn't be needed at all. | |
| 192 | LIB_DEPENDS should come before USES: https://docs.freebsd.org/en/books/porters-handbook/order/#porting-order | |
| 199 | I didn't even think about that... So all other Haskell ports we have that depend in libyaml aren't using the system library. We should probably sink constraints such as these down to Uses/cabal.mk to make them affect all the ports. But that's for later, of course. | |
| 212 | Both these knobs could be removed, CABAL_EXECUTABLES is set to ${PORTNAME} by default. | |
Hi. I had a quick look into the current version, and I still see lots of issues with the Makefile (as reported by portlint, portclippy, and portfmt).
Also, currently I am not able to access my own server, which means that I cannot try out this version in poudriere;
for now, I tried in a local arm64 machine (which comes with ghc-9.10.3, which should work for this ghcup version), and it errors out during compilation... like this:
Error: [Cabal-7107]
Could not resolve dependencies:
[0] trying: Cabal-3.16.1.0 (user goal)
[1] trying: unix-2.8.7.0/installed-inplace (dependency of Cabal)
[2] next goal: os-string (user goal)
[2] rejecting: os-string-2.0.10 (conflict: unix => os-string==2.0.7/installed-inplace)
[2] rejecting: os-string-2.0.7/installed-inplace (constraint from user target requires ==2.0.10)
[2] fail (backjumping, conflict set: os-string, unix)
After searching the rest of the dependency tree exhaustively, these were the goals I've had most trouble fulfilling: os-string, Cabal, unix
Small update:
I fixed the sorting, and all the findings of the linter tools, as specified in the previous comment.
Also removed os-string from the dependencies, and it then compiles on my local machine.
Did not try with poudriere yet, through -- will be able to do this maybe this Sat evening or Sun.
With those changes in place (and assuming poudriere is happy), then it LGTM.
Will commit the changes ASAP.
BTW - hasufell_posteo.de - thanks a lot for your effort - this is very much appreciated :-)
Ah right... it's possible that it picked up my local 9.6.7 compiler when I was running make manually in the ports tree. I'll sync it up with 9.10.3. The os-string issue is a result of that, since 9.6.7 does not ship with it.
Checked with poudriere and all looks fine.
Also checked microlens dependency, as per comment from arrowd@ - seems we really need this, otherwise compilation breaks.
Thanks!