I was too focused on getting through the meeting to express my 
overwhelming joy at getting to the bottom of this issue -- thanks for the 
tireless work in the face of very confusing behavior, Mike.

(I've also now determined that I was not the culprit which only adds to my 
joy.  And I can clear Steve's poor name as well, thankfully).

-Brad


On Tue, 16 Sep 2014, Michael Noakes wrote:

>  Branch: refs/heads/master
>  Home:   https://github.com/chapel-lang/chapel
>  Commit: b67945810fe5c2573e9a002358f17c1a8e37b02f
>      
> https://github.com/chapel-lang/chapel/commit/b67945810fe5c2573e9a002358f17c1a8e37b02f
>  Author: Michael Noakes <[email protected]>
>  Date:   2014-09-15 (Mon, 15 Sep 2014)
>
>  Changed paths:
>    M compiler/include/map.h
>
>  Log Message:
>  -----------
>  Fix for a pointer truncation bug that is surfaced on Mac OS-X
>
>
>  Commit: 0022cbdbacf77e0e7c253230a932d84ebb63c5fc
>      
> https://github.com/chapel-lang/chapel/commit/0022cbdbacf77e0e7c253230a932d84ebb63c5fc
>  Author: Michael Noakes <[email protected]>
>  Date:   2014-09-16 (Tue, 16 Sep 2014)
>
>  Changed paths:
>    M compiler/include/map.h
>
>  Log Message:
>  -----------
>  Merge pull request #436 from noakesmichael/trunc-bug
>
> Fix for a pointer truncation bug that is surfaced on Mac OS-X
>
> This one line change addresses a long standing problem with managing Maps 
> that has been exposed as intermittent failures on Mac OS X.  There was an 
> operation on a MapElem included an explicit cast from uintptr_t to unsigned 
> int hence losing the upper 32 bits of the address.  In this new version we 
> simply preserve the uintptr_t.
>
> My small concern is that I am unsure "why" the original author defined the 
> return type to be a unsigned int in the first place.
>
> a) This code has been tested against a repeatable failure case on Mac.
>
> b) This code ran Lulesh for 2500 iterations; it had been failing 
> approximately every 100 iterations or so.
>
> c) Elliot has run this version on a program he has that used to fail 
> intermittently.
>
> d) Passed paratest.
>
> e) Compiled on Mac/clang and Linux/gnu
>
> Reviewed by Elliot
>
>
> Compare: 
> https://github.com/chapel-lang/chapel/compare/e7636e09e28e...0022cbdbacf7
>

------------------------------------------------------------------------------
Want excitement?
Manually upgrade your production database.
When you want reliability, choose Perforce.
Perforce version control. Predictably reliable.
http://pubads.g.doubleclick.net/gampad/clk?id=157508191&iu=/4140/ostg.clktrk
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers

Reply via email to