Stacktrace not working on the Windows 64/32 binaries

[jfrolich]

There is no stacktrace present in the latest windows binary builds (6 days ago).

Version 0.3.0-prerelease+1576 (2014-02-14 22:18 UTC)
Commit 494a451* (5 days old master)
i686-w64-mingw32

Version 0.3.0-prerelease+1576 (2014-02-14 22:18 UTC)
Commit 494a451* (5 days old master)
i686-w64-mingw32

Showcase

julia> one(e) = e?error("crash here"):nothing
one (generic function with 1 method)

julia> two(e) = e?one(e):nothing
two (generic function with 1 method)

julia> three(e) = e?two(e):nothing
three (generic function with 1 method)

julia> three(true)
ERROR: crash here

Related Issue: #3031 (This issue is closed).

---

one(e) = e?error("crash here"):nothing
two(e) = e?one(e):nothing
three(e) = e?two(e):nothing
three(true)

[tkelman]

+1, having same problem.

[jfrolich]

Hopefully this bug can be squashed, because it is no fun coding without a stacktrace :).

Besides that: Loving Julia so far. I am building a decision tree algorithm that supports (optimizes for) sparse matrices. Useful for text classification.

[kmsquire]

Sorry, hit the close button by accident.

Off-topic, but @jfrolich, if your code might be generally useful to others, you might consider adding it to the DecisionTrees package.

[jfrolich]

@kmsquire Certainly!

[jfrolich]

Does anyone know when a new nightly will be built for windows?

[jfrolich]

This is still not fixed in the newest nightly released 5 hours ago. Think this should be an important bug-fix before the release of 0.3. (I miss my MacBook, but unfortunately my company only lets me use Windows :()

[vtjnash]

somewhat unrelated, but I really like the work the LDC guys are going: http://redstar.de/ldc/win64eh_all_20140201.diff

hopefully this makes it into LLVM 3.5 so we can jump to that and improve error handling on windows

[vtjnash]

funny enough, I wrote the probable solution to this problem in #3031 when I closed it, then promptly forgot. ok, so nobody's laughing, at least the fix isn't too hard.