Various fixes - #1
Open
angerman wants to merge 57 commits into
Open
Conversation
(cherry picked from commit 650ebfc2363ced3e102b7a32f2266655e31c4593)
(cherry picked from commit 4c58525d98c1b09b41a9dcba02e616987a8fea2e)
(cherry picked from commit 50d0abb3317ac439a4e7495b185a64af9b7b9300)
(cherry picked from commit 5a545fd94dc19cb1ec1853ecb9bd313e56f34516)
(cherry picked from commit b3a865acd224e2773e3f16caceb83ad9b4100fc3)
(cherry picked from commit 9151db2a9a61d7f5fe52ff8836f18bbd0fd8933c)
(cherry picked from commit 43a979272d9addc29fbffc2e8542c5d96e993d73)
We only have this for windows in haskell.nix so far 😅
Without these, we'll end up with lots of > `getDirectoryContents:openDirStream: invalid argument (Invalid argument)` this is due to malloc failing (the malloc from the in-memory loaded libc). loading two libcs into the same process is never a good idea.
This is used by readCreateProcessWithExitCode.
Is this still true? It looks like @alexfmpe makes use of this PR in NixOS/nixpkgs#445672, but I don't see any additional GHC patches added there. What's missing to actually merge this? |
|
I think that refers to Note it's not applied for 9.12 Not sure about the other part
|
Fix iserv-proxy build on windows
TCP networking inside nix-linux-builder VMs is unreliable, causing iserv-proxy to hang or fail when connecting to the interpreter via sockets. This adds two new modes: - Interpreter --stdio: reads/writes protocol data on stdin/stdout instead of TCP. Uses hDuplicate/hDuplicateTo to redirect the Haskell stdout handle to stderr, preventing library code from corrupting the data channel. - Proxy --pipe <cmd>: spawns the interpreter as a subprocess and communicates via its stdin/stdout pipes, bypassing TCP entirely. Both modes are opt-in; the original TCP mode is unchanged. Also redirects all diagnostic output (trace, hook messages) to stderr in both old and new code paths.
The dynamic linker fails to load this library when it sees
unresolved references via the `SYM(...)` table. Provide
`MISSING_FUN` stubs for the symbols our current armv7a NDK's
bionic / libc++ don't ship; they're never expected to be
invoked at runtime — if any actually does, the stub exits with
a diagnostic.
Symbols covered: posix_spawn{,attr}_{init,destroy,setflags,
setsigdefault,addopen,adddup2,addclose,addchdir_np},
__cxa_guard_{abort,acquire,release}, __cxa_thread_finalize,
__libc_{stdio_cleanup,shared_globals}, getentropy, statx,
__fsetxattr, __arm_fadvise64_64.
A second pass surfaced this one too — same rationale as the previous commit.
…_assert mingw-w64 14.0.0 added an `_assert` member to `libmingwex.a` (13.0.0 had none, so `_assert` came from the UCRT import library). That member calls `fileno` and imports `__msvcrt_assert`. wine's ucrtbase implements both but only exports the underscore spellings, so `GetProcAddress` returns NULL for the aliases. GHC's runtime linker loads every member of an archive, so the unresolvable `fileno` fails all of `libmingwex.a`, and template haskell then dies with libmingwex.a: unknown symbol `fileno' HSghc-bignum-...-inplace.o: unknown symbol `_assert' <no location info>: error: unable to load unit `ghc-bignum-1.3' `strncasecmp` is the same story one level up, via unix-time's object. The extra symbol table is consulted before objects and DLLs and is in place before anything is loaded, so it is the right place for these: a `-l` archive is loaded only after `initLoaderState` has already resolved the command-line packages. `__imp___msvcrt_assert` gets an indirection entry (the shape of GHC's Note [_iob_func symbol]) pointing at a local implementation rather than at `_assert`, which is the caller here and would recurse. Verified on x86_64-linux with `--builders ""`: haskell.nix's ucrt64 th-dlls-minimal (all six variants) against GHC 9.6.7, 9.8.4, 9.10.3, 9.12.4 and 9.14.1, plus ucrt64 th-dlls, all of which the mingw-w64 14 bump had broken. Claude-Session: https://claude.ai/code/session_015iMn9nFmfb5DPqqW5a12co
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This restores most of the iserv functionality. It also crucially adds support for iserv-symbols. This requires a GHC patch, and will only work with haskell.nix.