Segfault after using workspace()

[micahjsmith]

I get a segfault with the following commands in the REPL:

using Blosc
workspace()
using Blosc
exit()

The full error output can be seen here: https://gist.github.com/micahjsmith/60b4556d67c47665ba587b11afa1d6b7

See also:

julia> versioninfo()
Julia Version 0.4.5
Commit 2ac304d* (2016-03-18 00:58 UTC)
Platform Info:
  System: Linux (x86_64-redhat-linux)
  CPU: Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.3

[stevengj]

Looking at the Blosc source code, it seems like it is not safe to call blosc_init twice, which is what will happen here (__init__ is called twice).

One thing we could do instead is to call blosc_set_nthreads(blosc_get_nthreads()), which from the source automatically calls blosc_init if it hasn't been initialized yet, and otherwise does nothing.

Another option might be to just not call blosc_init at all, since it looks from the source like blosc_init() is lazily called by any routine that needs it, but I'm wary of this because the Blosc documentation says you are supposed to call blosc_init. Even the set_nthreads trick seems iffy according to Blosc/c-blosc#150, but since g_initlib is a static (unexported) variable I don't see much choice.

[FrancescAlted]

Perhaps Blosc/c-blosc@5905212 would help?

[stevengj]

@FrancescAlted, I think you also need a if (g_initlib) return; check in blosc_init itself to prevent blosc_init from doing anything if it is called twice.

[FrancescAlted]

Good point. Fixed in Blosc/c-blosc@4b271d1.

[stevengj]

Great, thanks!

[stevengj]

Should be fixed on the latest master by the bump to c-blosc 1.11.1