| 26 Apr 2025 |
 aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | enableUnregisterised | 09:35:31 |
 Alex | In configurePhase? Might need a bit of patching then :/ | 09:35:59 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | Is cross broken in newer versions? | 09:36:26 |
| Alex | Last I checked, yes. | 09:36:44 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @aleksana:mozilla.org
Both results in
/nix/store/yfnsdzxp90xkwaxnm467y660nsj4zzgb-libffi-loongarch64-unknown-linux-gnu-3.4.7-dev/include/ffitarget.h:38:2: error: #error "libffi was configured for a LoongArch target but this does not appear to be a LoongArch compiler."
38 | #error \
| ^~~~~
/nix/store/yfnsdzxp90xkwaxnm467y660nsj4zzgb-libffi-loongarch64-unknown-linux-gnu-3.4.7-dev/include/ffitarget.h:66:2: error: #error unsupported LoongArch base architecture
66 | #error unsupported LoongArch base architecture
| ^~~~~
In file included from _build/stage0/libraries/ghci/build/GHCi/FFI_hsc_make.c:1:
FFI.hsc: In function ‘main’:
FFI.hsc:171:16: error: ‘FFI_DEFAULT_ABI’ undeclared (first use in this function); did you mean ‘FFI_LAST_ABI’?
/build/ghc-9.8.4-source/_build/stageBoot/lib/template-hsc.h:45:10: note: in definition of macro ‘hsc_const’
45 | if ((x) < 0) \
| ^
FFI.hsc:171:16: note: each undeclared identifier is reported only once for each function it appears in
/build/ghc-9.8.4-source/_build/stageBoot/lib/template-hsc.h:45:10: note: in definition of macro ‘hsc_const’
45 | if ((x) < 0) \
| ^
Is this the error or there are more? | 09:37:07 |
| Alex | There are more to come. | 09:37:28 |
| Alex | pkgsCross.loongarch64-linux.haskell.compilers.native-bignum.ghc96 should show you those errors, provided we don't have an assert there to save users time. | 09:38:31 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | There's a 9.6.6 deb for loongarch64 in debian ports actually | 09:39:49 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | But it's 140M and I don't know how to make it persistent | 09:40:13 |
| Alex | If you can run it check ghc --info.
It'll give some insight into how they built it. Most likely native unregjsterised. | 09:41:08 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @alex:tunstall.xyz
If you can run it check ghc --info.
It'll give some insight into how they built it. Most likely native unregjsterised. They do support llvm | 09:50:12 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | The problem is in our derivation I think, some guide just said they cross compiled ghc 9.6 like your suggestion without issues | 09:51:03 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @aleksana:mozilla.org
Is because of this
#ifndef __loongarch__
#error \
"libffi was configured for a LoongArch target but this does not appear to be a LoongArch compiler."
#endif
Because _build/stageBoot/bin/loongarch64-unknown-linux-gnu-hsc2hs --cc=gcc --ld=gcc | 09:58:26 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @aleksana:mozilla.org
I have to use pkgsCross.loongarch64-linux.haskell.packages.native-bignum.ghc984.ghc and temporarily override enableUnregisterised So there's apparently no native-bignum for newer ghc, and even if native-bignum is prefixed, it is still hadrian | 10:29:53 |
 sterni (he/him) | aleksana 🏳️⚧️ (force me to bed after 18:00 UTC): using foreign packages in stead of a bindist failed last time someone tried, I think GHC has a lot of absolute path references that are a big pain to patch after the fact | 11:24:50 |
| sterni (he/him) | aleksana 🏳️⚧️ (force me to bed after 18:00 UTC): native-bignum is just GHC without gmp, no relation to the build system | 11:25:07 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @sternenseemann:systemli.org
aleksana 🏳️⚧️ (force me to bed after 18:00 UTC): native-bignum is just GHC without gmp, no relation to the build system I don't get it why native bignum is an option alongside hadrian | 11:26:11 |
| sterni (he/him) | it's not along side hadrian at all | 11:26:50 |
| sterni (he/him) | native-bignum is the name for the bignum library GHC has which is native, i.e. written in Haskell | 11:27:16 |
| sterni (he/him) | it can be used as an alternative to gmp which is the default | 11:27:28 |
| sterni (he/him) | hadrian is the build system of GHC which was made the default in 9.4 and exclusive in 9.6 | 11:27:46 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @sternenseemann:systemli.org
it's not along side hadrian at all But you see every version is either defined with import common-hadrian.nix or import common-make-native-bignum.nix | 11:27:54 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | What does this mean | 11:28:17 |
| sterni (he/him) | it's named like that because alongside common-make-native-bignum.nix there would be common-make-integer-simple.nix, but since we only have one GHC that would use it (8.10.7.nix) I never bothered to make a common file | 11:29:17 |
| sterni (he/him) | and they are basically the same, they just renamed integer-simple to native-bignum at some point, so some inputs in the nix expression change which makes it annoying to make a common expression for both | 11:30:01 |
| sterni (he/him) | (it would be kind of a good project to make some cross compiled GHCs and upload them to tarballs.nixos.org before 25.11, doesn't look like we can keep around GHC 9.4 for that long) | 11:32:32 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @sternenseemann:systemli.org
(it would be kind of a good project to make some cross compiled GHCs and upload them to tarballs.nixos.org before 25.11, doesn't look like we can keep around GHC 9.4 for that long) But the current expression is a bunch of spaghetti and I don't think I can jump across all the cases as I'm not the writer of that | 11:35:30 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | Plus I don't understand haskell yet | 11:35:45 |
| sterni (he/him) | Yeah, I did not mean you should do it. | 11:36:01 |
 Alyssa Ross | the most pragmatic approach might be to focus on other loongarch64 things for now and hope upstream has bootstrap binaries before too long. | 11:36:32 |
