lua: fix on darwin by using makeBinaryWrapper #172749

zaphar · 2022-05-12T20:03:12Z

Darwin execve has issues with shebang lines that point to other scripts
with shebang lines. A new makeBinaryWrapper hook was added to workaround
the issue on darwin. See #171473 and #23018 for more information. This
uses that binary wrapper to fix packages like sile.

I'm not sure this can be considered complete but it appears to work for
sile at least.

cc: @alerque

zaphar · 2022-05-12T20:06:29Z

This cannot be considered complete yet but it appears to work for
sile at least. There is an issue in the checkPhase that has something to do with libtexpdf that seems unrelated but I'm still digging into it.

/nix/store/kmhw99yxbwgqai9v8b3rwb4nf31z1cml-lua-5.2.4/bin/lua: error loading module 'justenoughlibtexpdf' from file './core/justenoughlibtexpdf.so':
	dlopen(./core/justenoughlibtexpdf.so, 0x0006): Library not loaded: /nix/store/q97r3hiic0jg6ks95xylazf9m21hd1wh-sile-0.12.5/lib/libtexpdf.0.dylib
  Referenced from: /Users/zaphar/projects/nixpkgs/sile-0.12.5/core/justenoughlibtexpdf.so
  Reason: tried: '/nix/store/q97r3hiic0jg6ks95xylazf9m21hd1wh-sile-0.12.5/lib/libtexpdf.0.dylib' (no such file), '/usr/local/lib/libtexpdf.0.dylib' (no such file), '/usr/lib/libtexpdf.0.dylib' (no such file)
stack traceback:
	[C]: in ?
	[C]: in function 'require'
	./core/libtexpdf-output.lua:1: in main chunk
	[C]: in function 'require'
	./core/sile.lua:78: in function 'init'
	./sile:44: in main chunk
	[C]: in ?
Syntax Error: Document stream is empty
make: *** [Makefile:1381: selfcheck] Error 1

cc: @alerque in case he knows the cause of the above.

zaphar · 2022-05-12T20:29:48Z

Another issue I'm working through is this one:

$> NIXPKGS_ALLOW_BROKEN=1 nix run --impure .#sile -- --help                                                                                                                                                                ~/projects/nixpkgs

error: builder for '/nix/store/d5s9q2hgpxp3qhwh4qys8mq344a46rrf-lua5.4-compat53-0.7-1.drv' failed with exit code 1;
       last 10 log lines:
       > updateAutotoolsGnuConfigScriptsPhase
       > configuring
       > building
       > installing
       > Missing dependencies for compat53 0.7-1:
       >    lua >= 5.1, < 5.4 (not installed)
       >
       > compat53 0.7-1 depends on lua >= 5.1, < 5.4 (not installed)
       >
       > Error: Could not satisfy dependency lua >= 5.1, < 5.4: Rock lua 5.4-1 is already provided by VM or via 'rocks_provided' in the config file.
       For full logs, run 'nix log /nix/store/d5s9q2hgpxp3qhwh4qys8mq344a46rrf-lua5.4-compat53-0.7-1.drv'.
error: 1 dependencies of derivation '/nix/store/0fykn71wjj4jk2mqjcg4khkwxjidpabq-lua-5.4.3-env.drv' failed to build
error: 1 dependencies of derivation '/nix/store/rg8m4ccf4yzb0i4rr4zp1hqypk64vqvs-sile-0.12.5.drv' failed to build

It appears to be unrelated to my change specifically and is perhaps due the compat version being incompatible with the lua version.

pkgs/top-level/all-packages.nix

zaphar · 2022-05-12T20:55:51Z

Running luacheck succeeds with this change.

$> nix run --impure .#lua53Packages.luacheck -- --help                                                                                                                                              ~/projects/nixpkgs
Usage: luacheck ([--config <config>] | [--no-config])
       ([--default-config <default_config>] | [--no-default-config])
       [-h] [-g] [-u] [-r] [-a] [-s] [--no-self] [--std <std>] [-c]
       [-d] [-t] [-m] [--max-line-length <length>]
       [--no-max-line-length] [--max-code-line-length <length>]
       [--no-max-code-line-length] [--max-string-line-length <length>]
       [--no-max-string-line-length]
       [--max-comment-line-length <length>]
      ....

thiagokokada

There is an unrelated cargo-feature commit here.

zaphar · 2022-05-12T21:08:52Z

Whoops sorry about that cargo-feature commit. I'll remove it. I think a rebase went wrong on my end.

zaphar · 2022-05-12T21:10:33Z

The only other package I can find with lua as a dependency that is also marked broken on darwin is pdns but it is marked broken on darwin anyway since it requires systemd.

zaphar · 2022-05-12T21:37:59Z

The only remaining issue I have to work through is the libtexpdf problem when the checkPhase runs from above. But that's about all the time I have for today. I'll get back to it tomorrow if I can. @alerque , if you know what that is happening let me know.

doronbehar · 2022-05-13T09:07:23Z

Quoting from #172341 :

I think it was marked as broken on Darwin due to wrapper issues (#171473), but it shouldn't be in general. I don't know what that is about.

@alerque I see now that lua5.4-zlib was broken and that's what caused sile to be marked as broken by nixpkgs-review.

As for the compat53 issue, perhaps @alerque you'd wish to open a new PR with this diff:

diff --git c/pkgs/tools/typesetting/sile/default.nix i/pkgs/tools/typesetting/sile/default.nix
index 4bff8ab3ae3..5e83cdd766f 100644
--- c/pkgs/tools/typesetting/sile/default.nix
+++ i/pkgs/tools/typesetting/sile/default.nix
@@ -18,7 +18,6 @@ let
   luaEnv = lua.withPackages(ps: with ps; [
     cassowary
     cosmo
-    compat53
     linenoise
     lpeg
     lua-zlib
@@ -33,6 +32,8 @@ let
     penlight
     stdlib
     vstruct
+  ] ++ lib.optionals (lib.versionOlder lua.luaversion "5.3") [
+    compat53
   ]);
 in

doronbehar

Diff looks pretty good, besides the libtexpdf issue and the doCheck now disabled on sile. cc @teto for a 2nd eye

pkgs/tools/typesetting/sile/default.nix

pkgs/development/interpreters/lua-5/interpreter.nix

zaphar · 2022-05-13T19:10:35Z

I keep getting bogged down in figuring out the libtexpdf and compat53 stuff. I'm going to exclude them from this change for now. I can workaround the libtexpdf thing with a doCheck override for the moment and the compat53 thing should be excluded from this change anyway I think.

zaphar · 2022-05-16T18:49:42Z

I've made the changes to leave the makeWrapper interface in place and tested the same way as above. I'll note that this diverges from the way python and ruby make this change but I don't think that's necessarily problematic. It just means lua is setup slightly different from the other interpreters.

vcunat · 2022-05-21T19:51:05Z

Lua stuff now got very broken on aarch64-darwin, apparently. And this commit is the main suspect chosen from the range where it started. Log: https://hydra.nixos.org/build/177341824 PR: #173605.

zaphar · 2022-05-21T22:06:02Z

That log doesn't give much to go on. Is it just timing out or is something else going on?

vcunat · 2022-05-22T07:11:58Z

I don't know, I don't have any compatible machine. I at least forced Hydra to retry a few times, but it always failed.

zaphar · 2022-05-22T12:07:48Z

I have a compatible machine. I'll try to figure it out but it will be a couple days before I can get to it. I understand if this has to be reverted in the meantime.

…

On Sun, May 22, 2022 at 3:12 AM Vladimír Čunát ***@***.***> wrote: I don't know, I don't have any compatible machine. I at least forced Hydra to retry a few times, but it always failed. — Reply to this email directly, view it on GitHub <#172749 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAJ5NYGQ5LVHCY3YS3MV73VLHM4XANCNFSM5VZKWAZQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Jeremy Wall http://jeremy.marzhillstudios.com ***@***.***

This reverts commit 92f4c6e. On aarch64-darwin this completely broke lua instead of improving it; let's revert at least until that's resolved. #172749 (comment)

zaphar · 2022-05-22T17:59:20Z

I was not able to replicate this on my my fix branch but I was able to replicate on staging. If it is this change then it's an interaction between this change and the other changes on staging. I may not be able to dig in any more till next week sometime though.

vcunat · 2022-05-22T18:16:44Z

I noticed in the meantime that also some coqPackages.* also now regressed in a similar way; example: https://hydra.nixos.org/build/177668313 Still, this revert really did avoid the mass lua failure.

zaphar · 2022-05-22T18:38:51Z

Digging in a little more it would appear that the lua binary is what is getting killed with prejudice. It's not clear why to me yet. Nor why it works in my branch but not staging. At a first guess perhaps something changed with the binaryWrapper that is causing it to get killed?

zaphar · 2022-05-22T18:47:30Z

Here is why lua is getting killed. It looks like a codesigning issue with the binaryWrapper that was introduced in staging.

Translated Report (Full Report Below)
-------------------------------------

Incident Identifier: D3F22F82-1FD7-4F98-83C9-84216E3802B4
CrashReporter Key:   E2FFED06-E6C8-B5A0-2FB3-130A4763A6F2
Hardware Model:      MacBookPro17,1
Process:             lua [67977]
Path:                /Volumes/VOLUME/*/lua
Identifier:          lua
Version:             ???
Code Type:           ARM-64 (Native)
Role:                Unspecified
Parent Process:      bash [67945]
Coalition:           org.nixos.nix-daemon [79]
Responsible Process: nix [127]

Date/Time:           2022-05-17 21:10:01.4506 -0400
Launch Time:         2022-05-17 21:10:01.3506 -0400
OS Version:          macOS 12.3.1 (21E258)
Release Type:        User
Report Version:      104

Exception Type:  EXC_BAD_ACCESS (SIGKILL (Code Signature Invalid))
Exception Subtype: UNKNOWN_0x32 at 0x0000000100d64000
Exception Codes: 0x0000000000000032, 0x0000000100d64000
VM Region Info: 0x100d64000 is in 0x100d64000-0x100d6c000;  bytes after start: 0  bytes before end: 32767
      REGION TYPE                    START - END         [ VSIZE] PRT/MAX SHRMOD  REGION DETAIL
      UNUSED SPACE AT START
--->  mapped file                 100d64000-100d6c000    [   32K] r-x/r-x SM=COW  ...t_id=a65ec1eb
      mapped file                 100d6c000-100d70000    [   16K] rw-/rw- SM=COW  ...t_id=a65ec1eb
Exception Note:  EXC_CORPSE_NOTIFY
Termination Reason: CODESIGNING 2 

Highlighted by Thread:  0

Backtrace not available

No thread state (register information) available

Binary Images:
               0x0 - 0xffffffffffffffff ??? (*) <00000000-0000-0000-0000-000000000000> ???

Error Formulating Crash Report:
_dyld_process_info_create failed with 6
dyld_process_snapshot_get_shared_cache failed
Failed to create CSSymbolicatorRef - corpse still valid ¯\_(ツ)_/¯
thread_get_state(PAGEIN) returned 0x10000003: (ipc/send) invalid destination port
thread_get_state(EXCEPTION) returned 0x10000003: (ipc/send) invalid destination port
thread_get_state(FLAVOR) returned 0x10000003: (ipc/send) invalid destination port

EOF

-----------
Full Report
-----------

{"app_name":"lua","timestamp":"2022-05-17 21:10:08.00 -0400","app_version":"","slice_uuid":"00000000-0000-0000-0000-000000000000","build_version":"","platform":0,"share_with_app_devs":0,"is_first_party":1,"bug_type":"309","os_version":"macOS 12.3.1 (21E258)","incident_id":"D3F22F82-1FD7-4F98-83C9-84216E3802B4","name":"lua"}
{
  "uptime" : 130000,
  "procLaunch" : "2022-05-17 21:10:01.3506 -0400",
  "procRole" : "Unspecified",
  "version" : 2,
  "userID" : 302,
  "deployVersion" : 210,
  "modelCode" : "MacBookPro17,1",
  "procStartAbsTime" : 3314279617860,
  "coalitionID" : 79,
  "osVersion" : {
    "train" : "macOS 12.3.1",
    "build" : "21E258",
    "releaseType" : "User"
  },
  "captureTime" : "2022-05-17 21:10:01.4506 -0400",
  "incident" : "D3F22F82-1FD7-4F98-83C9-84216E3802B4",
  "bug_type" : "309",
  "pid" : 67977,
  "procExitAbsTime" : 3314282012480,
  "translated" : false,
  "cpuType" : "ARM-64",
  "procName" : "lua",
  "procPath" : "\/Volumes\/VOLUME\/*\/lua",
  "parentProc" : "bash",
  "parentPid" : 67945,
  "coalitionName" : "org.nixos.nix-daemon",
  "crashReporterKey" : "E2FFED06-E6C8-B5A0-2FB3-130A4763A6F2",
  "responsiblePid" : 127,
  "responsibleProc" : "nix",
  "wakeTime" : 1722,
  "sleepWakeUUID" : "4FBB9758-0D22-4F9E-B8B4-9F9324F63C59",
  "sip" : "enabled",
  "vmRegionInfo" : "0x100d64000 is in 0x100d64000-0x100d6c000;  bytes after start: 0  bytes before end: 32767\n      REGION TYPE                    START - END         [ VSIZE] PRT\/MAX SHRMOD  REGION DETAIL\n      UNUSED SPACE AT START\n--->  mapped file                 100d64000-100d6c000    [   32K] r-x\/r-x SM=COW  ...t_id=a65ec1eb\n      mapped file                 100d6c000-100d70000    [   16K] rw-\/rw- SM=COW  ...t_id=a65ec1eb",
  "isCorpse" : 1,
  "exception" : {"codes":"0x0000000000000032, 0x0000000100d64000","rawCodes":[50,4309008384],"type":"EXC_BAD_ACCESS","signal":"SIGKILL (Code Signature Invalid)","subtype":"UNKNOWN_0x32 at 0x0000000100d64000"},
  "termination" : {"namespace":"CODESIGNING","flags":0,"code":2},
  "vmregioninfo" : "0x100d64000 is in 0x100d64000-0x100d6c000;  bytes after start: 0  bytes before end: 32767\n      REGION TYPE                    START - END         [ VSIZE] PRT\/MAX SHRMOD  REGION DETAIL\n      UNUSED SPACE AT START\n--->  mapped file                 100d64000-100d6c000    [   32K] r-x\/r-x SM=COW  ...t_id=a65ec1eb\n      mapped file                 100d6c000-100d70000    [   16K] rw-\/rw- SM=COW  ...t_id=a65ec1eb",
  "extMods" : {"caller":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"system":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"targeted":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"warnings":0},
  "usedImages" : [
  {
    "size" : 0,
    "source" : "A",
    "base" : 0,
    "uuid" : "00000000-0000-0000-0000-000000000000"
  }
],
  "legacyInfo" : {
  "threadHighlighted" : 0
},
  "reportNotes" : [
  "_dyld_process_info_create failed with 6",
  "dyld_process_snapshot_get_shared_cache failed",
  "Failed to create CSSymbolicatorRef - corpse still valid ¯\\_(ツ)_\/¯",
  "thread_get_state(PAGEIN) returned 0x10000003: (ipc\/send) invalid destination port",
  "thread_get_state(EXCEPTION) returned 0x10000003: (ipc\/send) invalid destination port",
  "thread_get_state(FLAVOR) returned 0x10000003: (ipc\/send) invalid destination port"
]
}

zaphar · 2022-05-22T19:12:54Z

@ncfavier, Is it possible that your changes here: #172366 broke the binaryWrapper on some platforms?

ncfavier · 2022-05-22T19:18:27Z

If there's a culprit in that PR it's most certainly 8b79ef2 (#172366). Maybe cctools wasn't enough after all, though it's strange that the makeBinaryWrapper tests succeeded on aarch64-darwin in my test PR.

zaphar · 2022-05-22T19:25:32Z

If there's a culprit in that PR it's most certainly 8b79ef2 (#172366). Maybe cctools wasn't enough after all, though it's strange that the makeBinaryWrapper tests succeeded on aarch64-darwin in my test PR.

It seems to one of those things that is only showing up when both this, and that change both landed. Lua wasn't using the binaryWrapper until this was merged in. I don't know if the Coq tests referenced above are also related or not.

ncfavier · 2022-05-22T19:35:14Z

Seems like coq_makefile is indeed a binary wrapper.

ncfavier · 2022-05-22T19:51:36Z

@zaphar would you be able to test #174038 ?

zaphar · 2022-05-22T19:53:02Z

@ncfavier, I'll give it a try tomorrow.

ncfavier · 2022-05-22T22:12:06Z

I see what's happening now: stripping binaries invalidates the signature (see #105026 (comment)). This is supposed to be avoided by having a wrapped version of strip that recreates the signature, but that wrapped version lives in cc, and cctools contains the non-wrapped version.

Indeed, in a previous build of coq we can see this:

strip is /nix/store/qlicbgwj6mddbcdcfsqqp7g5fq3rphd1-clang-wrapper-11.1.0/bin/strip

whereas in the current build, it's

strip is /nix/store/bi8ya5scc3305x4wa4nay5gmvriq4pxv-cctools-port-949.0.1/bin/strip

I am now convinced that #174038 fixes this.

This reverts commit 9f4060c. After the previous merge commit this should work now.

vcunat · 2022-05-23T15:49:52Z

OK, looks good on Hydra: https://hydra.nixos.org/eval/1763210#tabs-now-succeed

* origin/staging-next: (62 commits) Re-Revert "lua: fix on darwin by using makeBinaryWrapper (#172749)" openldap: fix cross-compilation makeBinaryWrapper: fix codesign on aarch64-darwin python3Packages.ldap: fix linking with openldap 2.5+ Revert "lua: fix on darwin by using makeBinaryWrapper (#172749)" wine: enable parallel build again pkgsi686Linux.gdb: fix formatting for 32-bit systems gtk4: Fix incorrect merge nixos/openldap: use upstream unit defaults openldap: update maintainers openldap: 2.4.58 -> 2.6.2 Revert "Add mingwW64-llvm cross-system." lua: fix on darwin by using makeBinaryWrapper (#172749) python310Packages.python-mimeparse: execute tests pandas: fix darwin build gtk3: 3.24.33 -> 3.24.33-2022-03-11 gtk4: patch fixing g-c-c crashes e2fsprogs: patch for CVE-2022-1304 firefox-unwrapped: fix cross compilation rustc: expose correct llvmPackages for cross compile ...

zaphar · 2022-05-23T18:50:52Z

@vcunat, @teto, @ncfavier, @doronbehar, and @alerque , Thanks for the quick action on these. As a nix user on darwin I really appreciate the patience for my first contribution to nixpkgs.

github-actions bot added the 6.topic: lua label May 12, 2022

zaphar mentioned this pull request May 12, 2022

Fix sile wrapper issues on darwin. #172407

Closed

13 tasks

ofborg bot added the 6.topic: darwin Running or building packages on Darwin label May 12, 2022

ofborg bot requested a review from doronbehar May 12, 2022 20:32

ofborg bot added 10.rebuild-darwin: 501+ 10.rebuild-darwin: 501-1000 10.rebuild-linux: 501+ 10.rebuild-linux: 2501-5000 labels May 12, 2022

zaphar force-pushed the darwin-sile branch from 0e44c99 to 2798582 Compare May 12, 2022 20:46

zaphar changed the title ~~Fix sile wrapper issues on darwin.~~ lua: fix on darwin by using makeBinaryWrapper May 12, 2022

alerque reviewed May 12, 2022

View reviewed changes

pkgs/top-level/all-packages.nix Outdated Show resolved Hide resolved

thiagokokada reviewed May 12, 2022

View reviewed changes

zaphar force-pushed the darwin-sile branch from 2798582 to 17416ec Compare May 12, 2022 21:09

zaphar force-pushed the darwin-sile branch from 17416ec to 701b59c Compare May 12, 2022 21:31

zaphar force-pushed the darwin-sile branch from 701b59c to a352af2 Compare May 13, 2022 02:05

doronbehar reviewed May 13, 2022

View reviewed changes

SuperSandro2000 reviewed May 13, 2022

View reviewed changes

pkgs/tools/typesetting/sile/default.nix Outdated Show resolved Hide resolved

SuperSandro2000 reviewed May 13, 2022

View reviewed changes

pkgs/development/interpreters/lua-5/interpreter.nix Outdated Show resolved Hide resolved

zaphar force-pushed the darwin-sile branch 2 times, most recently from 9728b75 to eeb8082 Compare May 13, 2022 19:08

zaphar force-pushed the darwin-sile branch 2 times, most recently from ff6b011 to c3ce148 Compare May 13, 2022 19:32

zaphar force-pushed the darwin-sile branch from e760782 to b26024b Compare May 16, 2022 18:47

alerque approved these changes May 16, 2022

View reviewed changes

ofborg bot added 10.rebuild-linux: 1001-2500 and removed 10.rebuild-linux: 2501-5000 labels May 16, 2022

doronbehar merged commit 92f4c6e into NixOS:staging May 17, 2022

ncfavier mentioned this pull request May 22, 2022

makeBinaryWrapper: fix codesign on aarch64-darwin #174038

Merged

vcunat added a commit that referenced this pull request May 23, 2022

Re-Revert "lua: fix on darwin by using makeBinaryWrapper (#172749)"

4ddea71

This reverts commit 9f4060c. After the previous merge commit this should work now.

doronbehar mentioned this pull request Oct 4, 2022

sile: Drop broken flag on Darwin #189582

Closed

stephank mentioned this pull request Oct 4, 2022

lua: add back Darwin makeBinaryWrapper fix #194389

Merged

13 tasks

abathur mentioned this pull request May 24, 2023

haproxy: enable Lua on macOS #233891

Merged

12 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

lua: fix on darwin by using makeBinaryWrapper #172749

lua: fix on darwin by using makeBinaryWrapper #172749

zaphar commented May 12, 2022 •

edited

Loading

zaphar commented May 12, 2022

zaphar commented May 12, 2022

zaphar commented May 12, 2022

thiagokokada left a comment

zaphar commented May 12, 2022

zaphar commented May 12, 2022

zaphar commented May 12, 2022

doronbehar commented May 13, 2022

doronbehar left a comment

zaphar commented May 13, 2022

zaphar commented May 16, 2022

vcunat commented May 21, 2022

zaphar commented May 21, 2022

vcunat commented May 22, 2022

zaphar commented May 22, 2022 via email

zaphar commented May 22, 2022

vcunat commented May 22, 2022

zaphar commented May 22, 2022

zaphar commented May 22, 2022 •

edited

Loading

zaphar commented May 22, 2022

ncfavier commented May 22, 2022 •

edited

Loading

zaphar commented May 22, 2022 •

edited

Loading

ncfavier commented May 22, 2022

ncfavier commented May 22, 2022

zaphar commented May 22, 2022

ncfavier commented May 22, 2022

vcunat commented May 23, 2022

zaphar commented May 23, 2022

lua: fix on darwin by using makeBinaryWrapper #172749

lua: fix on darwin by using makeBinaryWrapper #172749

Conversation

zaphar commented May 12, 2022 • edited Loading

zaphar commented May 12, 2022

zaphar commented May 12, 2022

zaphar commented May 12, 2022

thiagokokada left a comment

Choose a reason for hiding this comment

zaphar commented May 12, 2022

zaphar commented May 12, 2022

zaphar commented May 12, 2022

doronbehar commented May 13, 2022

doronbehar left a comment

Choose a reason for hiding this comment

zaphar commented May 13, 2022

zaphar commented May 16, 2022

vcunat commented May 21, 2022

zaphar commented May 21, 2022

vcunat commented May 22, 2022

zaphar commented May 22, 2022 via email

zaphar commented May 22, 2022

vcunat commented May 22, 2022

zaphar commented May 22, 2022

zaphar commented May 22, 2022 • edited Loading

zaphar commented May 22, 2022

ncfavier commented May 22, 2022 • edited Loading

zaphar commented May 22, 2022 • edited Loading

ncfavier commented May 22, 2022

ncfavier commented May 22, 2022

zaphar commented May 22, 2022

ncfavier commented May 22, 2022

vcunat commented May 23, 2022

zaphar commented May 23, 2022

zaphar commented May 12, 2022 •

edited

Loading

zaphar commented May 22, 2022 •

edited

Loading

ncfavier commented May 22, 2022 •

edited

Loading

zaphar commented May 22, 2022 •

edited

Loading