This is the mail archive of the mailing list for the frysk project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: cvs head has lots of test failures

Quoting Tim Moore <>:

Hash: SHA1

Phil Muldoon wrote:
I reverted LinuxTask.hava from 1.32. The endian patch was causing a few
issues. Perhaps the ultimate issue is in getIsa() using an elfFile which
uses a file handle. The theory is these handles are being queued for
Garbage Collection along with the object when the object is destroyed,
but the system is running out of handles before the Garbage Collection
is happening. Hopefully I have not totally mangled that explanation
though that people worked through on irc today. ;)

The problem is that the file descriptor passed to libelf is *never*
closed, whether or not the Elf Java object is garbage collected. I've
added an explicit close() method for the Elf object so that getIsa() can
release the file descriptor right away, and of course added the code to
close the file descriptor. frysk-core TestRunner tests run now, but I'm
going to sleep in it before checking in my code.

That is just what I thought of. Can you post your code here?

BTW. This make me think of another checkin policy: before checking in code, we can post the diff on the mail-list, if nobody reject in a specified time (one day, two days, or longer), it can get in. Another option is for some people to review the code, but that will take some more time.

If the first option is enabled, we are willing to test the new code in a as-short-as-possible time. Although we are mainly focused on ppc64, we also care about all aspect of Frysk anyway. :-)

On Thu, 2006-08-10 at 09:06 -0400, Elena Zannoni wrote:

Mark Wielaard writes:
 >  > We really need an autobuilder to catch such things early.

We really need people to run make check before checking stuff in.
Please folks be considerate to your fellow coders...

Mea culpa for not running the tests more carefully, though in my defense the tests were broken at the time on x8664 :)

I guess Mea culpa is French's saying for saying sorry. If it is, we need say sorry here too. :-) These getIsa() code is from us anyway.

I am now requesting my side's people to carefully test all the code they write. If they need to check in, they need at least try TestRunner and make check under relevant directory to make sure there is no regression.

- Wu Zhou

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]