Let me add that the backtrace is not always the same.
If I run Pkg.update() but typing it into the julia realine shell rather 
than passing it as an execute argument I get this b.t.:
https://gist.github.com/cdsousa/a18439acba05a64d8476

On the other hand if I do exactly the same thing but using 
julia-debug-basic instead of julia-debug-readline then it doesn't segfaults.

It seems to be some dangling/wild pointer... no?

On Thursday, February 6, 2014 2:32:26 AM UTC, Cristóvão Duarte Sousa wrote:
>
> Hi,
>
> I'm using Julia compiled from git in Arch linux, and in the last weeks 
> I've been getting almost random segmentation faults, mainly, but not only, 
> when running Pkg.update().
>
> I've followed the procedures from 
> https://gist.github.com/staticfloat/6188418 (except for the Github issue 
> opening) and I'm posting here the gdb logs. All logs refer to starting 
> Julia with an empty .julia/v0.3 folder.
>
> I started by running versioninfo(true) before Pkg.update(), but somehow in 
> this case it doesn't segfaults; here is a gdb log copypast for that case:
> https://gist.github.com/cdsousa/9cfa5307ae221e72f2e2#file-julia_ok_log
>
> If I just run Pkg.update(), also with an empty .julia/v0.3, it segfaults:
> https://gist.github.com/cdsousa/9cfa5307ae221e72f2e2#file-julia_nok_log
> (the log is huge since there is a lot of llvm code in the middle)
>
> I've also seen segmentation faults when running other commands, and I even 
> have got BoundsErrors from Base functions in a few occasions.
>
> I've also discovered that by removing the precompiled sys.so, the 
> segfaults disappear:
>
> https://gist.github.com/cdsousa/9cfa5307ae221e72f2e2#file-julia_nosysso_ok_log
>
> I have no knowledge about how to further tackle this issue.
> What can I do next?
>
> Thanks,
> Cristóvão
>

Reply via email to