[CMake] Fwd: Dart Test seg faults

[David Cole]

If you look at the number of bytes passed to realloc as hex it's 
"59D5AE1C." Could that be a pointer value? Perhaps a call to realloc is 
mistakenly passing a pointer to a size_t instead of dereferencing it 
first? Do you NOT see it when you're NOT running a cron job because the 
pointer values are different (and perhaps a reasonable size for a 
realloc call in the non-cron job case)? I don't know; I'm still learning 
when it comes to Unix...

Is there a realloc in that last test or is it coming from the testing 
framework?

What happens if you remove just that last test from your list of tests? 
Does everything else work?

Just some ideas to trigger further discussion... Hopefully helpful!

David


Alex Brooks wrote:

>When my nightly cron job runs the tests, I get output like the following:
>
>** make NightlyTest
>Building Utility NightlyTest...
>Test Nightly
>        Finding Tests
>        Found 6 tests
>        Running Tests, one icon per test, 50 icons per line
>        . == Passed     - == Failed     * == NotRun
>        .....make: *** [NightlyTest] Segmentation fault
>
>or
>
>$ make NightlyTest
>Building Utility NightlyTest...
>Test Nightly
>        Finding Tests
>        Found 9 tests
>        Running Tests, one icon per test, 50 icons per line
>        . == Passed     - == Failed     * == NotRun
>        ........unable to realloc 1507175964 bytes
>make: *** [NightlyTest] Aborted
>
>
>I have only gotten the former once, but the latter is easy to reproduce.
>For some reason it only occurs when run from a cron job -- I can't
>reproduce it manually.
>
>It always occurs on that last test, which produces an awful lot of output
>on stdout, so I suspect it may be related to that.
>
>Any ideas?
>  
>