Re: Hydra Failure

[Mike Miller]

On Tue, Apr 19, 2016 at 16:59:32 -0700, Rik wrote:
> 4/19/16
> 
> Mike,
> 
> The latest Hydra failure was well captured.
> 
>   libinterp/parse-tree/oct-parse.in.yy-tst ....................terminate
> called after throwing an instance of 'std::length_error'
>   what():  basic_string::_S_create
> panic: Aborted -- stopping myself...
> attempting to save variables to 'octave-workspace'...
> save to 'octave-workspace' complete
> /nix/store/yka2g35h56sp9qai6vm4ilijy57h5mqb-bash-4.3-p30/bin/bash: line 1: 
> 2134 Aborted                
> /nix/store/yka2g35h56sp9qai6vm4ilijy57h5mqb-bash-4.3-p30/bin/bash
> ../run-octave --norc --silent --no-history
> /tmp/nix-build-octave-0pre.drv-0/octave-4.1.0+/test/fntests.m
> /tmp/nix-build-octave-0pre.drv-0/octave-4.1.0+/test
> 
> Is there any way to get an interactive session going on the Hydra VM using
> this exact binary that failed?  The file oct-parse.in.yy-tst is very short
> at just 70 lines, but we need to be able to play around and narrow down
> which test is causing the segfault.

Sorry, I don't think the contents of the build are saved in the event of
a failure (but I don't know for sure, only what I see on the hydra web
interface).

The only thing I see you can do with a failure is "Reproduce locally"
which involves installing a NixOS system and recreating the build with
all of the exact input states. Which may or may not error the same way.

FWIW, I've built in a 32-bit Debian environment and not yet been able to
reproduce this error.

-- 
mike