eigfact() of a symmetric matrix causes julia to exit

[pgawron]

Following example run in REPL

G=randn(2,2)
F=eigfact(Symmetric(G'*G),4,1)

causes Julia to exit with error:

** On entry to DSYEVR parameter number  8 had an illegal value

julia version 0.3.0-rc4+20

[andreasnoack]

I see the same error message, but Julia doesn't exit. Please give your versioninfo()?

[pgawron]

julia> versioninfo()
Julia Version 0.3.0-rc4+20
Commit 70f4091 (2014-08-14 19:18 UTC)
Platform Info:
System: Linux (x86_64-linux-gnu)
CPU: Intel(R) Core(TM)2 Duo CPU U9400 @ 1.40GHz
WORD_SIZE: 64
BLAS: libblas.so.3
LAPACK: liblapack.so.3
LIBM: libopenlibm
LLVM: libLLVM-3.3

[andreasnoack]

Fixed by 8e3b176. @pgawron thank you for reporting.

[ivarne]

These issues seems relevant for backporting to 0.3

@andreasnoackjensen do you know how to git cherry-pick?

[andreasnoack]

Soon, I do.

[andreasnoack]

Okay. Just tried without luck. I had planned to do other things this evening than exercising git, so would it be possible for you to either explain how it works or simply cherry pick the three commits to 0.3? Thanks.

[ivarne]

Totally agree. Learning git makes for a terrible evening. Knowing it makes you a wizard 😸

In principle it should be as easy as:

cd release-julia # or wherever you have a git checkout of the 0.3 release
git checkout release-0.3
git pull
git cherry-pick 8e3b176b255c694964974cff766fb1f51b0eabca
git cherry-pick 64e21a3550ce132b85df1208f3517aac1c3059e5
git cherry-pick e03700a8b2b4b7afc3b33735c966c0227d688981
make test
# and if the tests work:
git push

but I get merge conflicts on the first commit because of #7992. The whole linalg code is way too cryptic for me, so I won't just try to mess too much with it.

For extra points, you can do git commit --amend after each chery-pick, to add a note in the commit message that this is a backport and reference the original commit.

[andreasnoack]

Done with bab9636, but no cherry picking. The conflicts were too deep.

[ivarne]

Thanks, that was just in time for the release.

[pgawron]

Thanks to @andreasnoackjensen and @ivarne for fixing those bugs. That was fast.