crash when evaluating module #2718
Comments
|
same crash happens with master 5886bc5. I wanted to investigate deeper and followed the hacking section https://nixos.org/nix/manual/#chap-hacking yet: then I tried to run the inst/bin/nix-shell and got |
|
Here is a full backtrace, I hope that's helpful. |
|
On Tue, 12 Mar 2019 23:28:12 -0700, Matthieu Coudron ***@***.***> wrote:
Here is a full backtrace, I hope that's helpful.
```
#0 0x00000000004361c8 in boost::io::basic_altstringbuf<char, std::char_traits<char>, std::allocator<char> >::overflow(int) ()
#1 0x00007ffff68850f9 in std::basic_streambuf<char, std::char_traits<char> >::xsputn(char const*, long) ()
from /nix/store/qh3pn6d1a6m31kb385pzi7vagwcbwayq-gcc-7.3.0-lib/lib/libstdc++.so.6
#2 0x00007ffff6875824 in std::basic_ostream<char, std::char_traits<char> >& std::__ostream_insert<char, std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&, char const*, long) ()
from /nix/store/qh3pn6d1a6m31kb385pzi7vagwcbwayq-gcc-7.3.0-lib/lib/libstdc++.so.6
#3 0x00007ffff76b753a in nix::printValue(std::ostream&, std::set<nix::Value const*, std::less<nix::Value const*>, std::allocator<nix::Value const*> >&, nix::Value const&) () from /nix/store/j40zb6apb6pl2n1imrpxikr1g221p448-nix-2.3pre6645_5886bc5/lib/libnixexpr.so
Maybe the `std::set` needs to be using one of the bdwgc allocators
instead of `std::allocator` to avoid badness?
(`gc_allocator<Value const*>` or `traceable_allocator<Value const*>`,
not sure but I'd guess the latter if at all)
…-----
Unfortunately I can't reproduce the problem, mostly because the linked
expressions (and mptcpanalyzer's shell.nix) seem to require use of
an overlay to provide the actual expression for mptcpanalyzer--
everything else just takes it for granted that the base exists
and extends it one way or another :D.
If you're able to distill the triggering expressions a bit
that might be helpful, but even if not maybe a pointer to
how one might invoke the problematic evaluation?
------
The stacktrace points to crashing while evaluating some
`builtins.trace`, which I've noted in your tree might be called
as part of a warning system -- does that sound correct?
Any chance you're dumping some insanely large tree of expressions
into `builtins.trace` or something?
Just a thought.
Not that trace should crash regardless :).
|
|
when I noticed trace was involved, I disabled the call in and it stopped crashing. There should have been 8 packages , 15 at most. |
|
it happened again also because of a |
|
I can reproduce this issue on https://github.com/jtojnar/nixpkgs/tree/nix-segfault with following versions: $ nix run -f . nixStable -c nix-build --version
nix-build (Nix) 2.3.4
$ nix-build --version
nix-build (Nix) 2.4pre7346_5e7ccdc9when building the NixOS manual using Stable Nix shows the same traceback as OP: #0 0x000000000044c127 in boost::io::basic_altstringbuf<char, std::char_traits<char>, std::allocator<char> >::overflow(int) ()
No symbol table info available.
Backtrace stopped: Cannot access memory at address 0x7ffd07c6d308with unstable, it changes slightly: #0 0x00000000004554f7 in std::basic_streambuf<char, std::char_traits<char> >::sputc (__c=116 't',
this=0x7fffe62924c8)
at /nix/store/yrm128w6m29dcq2yg1rlgyq037vjnv0p-gcc-9.3.0/include/c++/9.3.0/streambuf:431
#1 boost::io::basic_altstringbuf<char, std::char_traits<char>, std::allocator<char> >::overflow (
this=0x7fffe62924c8, meta=116) at include/boost/format/alt_sstream_impl.hpp:290
Backtrace stopped: Cannot access memory at address 0x7fffe62906b8 |
|
I am still seeing this on current master. I'm encountering this after adding a trace on purpose, so removing the trace statement isn't a workaround. I was trying to trace-print a derivation. As a workaround, I was able to trace-print the .drvPath attribute of the derivation instead. |
|
I can also confirm this on both Nix unstable ( A minimal example to reproduce looks like this: My The error looks like this on both unstable and stable: The culprit seems to be: CCing @edolstra & @infinisil who may have an idea how to investigate this further or how to fix this :) |
|
I marked this as stale due to inactivity. → More info |
|
@Ma27 I just tried reproducing this with your example and could not (but I do get an evaluation error). |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I am trying to write some module that generates a (neo)vim config based on buildInputs (will create a proper python3Env etc).
The crash happens when running this nix-shell :
https://github.com/teto/mptcpanalyzer/blob/fda8ffb12976c3de31f537386b02eb9ded09ca9d/contrib/shell.nix
with this nixpkgs (nixos-unstable plus my WIP patches):
teto/nixpkgs@a3b1357
When I try to test it, nix crashes:
nix (Nix) 2.2
I fear the backtrace might not be very helpful as is but I don't hae yet a good workflow on nixpkgs for gdb (any hint/tip appreciated):
The text was updated successfully, but these errors were encountered: