On Thu, Nov 12, 2009 at 5:57 PM, Henry Yei <[email protected]> wrote: >> -----Original Message----- >> From: Garrett Cooper [mailto:[email protected]] >> Sent: Thursday, November 12, 2009 1:43 AM >> To: Henry Yei >> Cc: [email protected]; LTP Mailing List >> Subject: Re: [LTP] [PATCH][ppoll01] fix testcase 1, 5: pass correct >> valuefor sigset_t for mips >> >> >> -----Original Message----- >> >> From: Garrett Cooper [mailto:[email protected]] >> >> Sent: Tuesday, November 10, 2009 10:13 PM >> >> To: Henry Yei >> >> Cc: [email protected]; LTP Mailing List >> >> Subject: Re: [LTP] [PATCH][ppoll01] fix testcase 1, 5: pass correct >> >> valuefor sigset_t for mips >> >> >> >> On Tue, Nov 10, 2009 at 5:14 PM, Henry Yei <[email protected]> wrote: >> >> >> -----Original Message----- >> >> >> From: Garrett Cooper [mailto:[email protected]] >> >> >> Sent: Monday, November 09, 2009 11:17 PM >> >> >> To: Henry Yei >> >> >> Cc: [email protected]; LTP Mailing List >> >> >> Subject: Re: [LTP] [PATCH][ppoll01] fix testcase 1, 5: pass >> correct >> >> >> valuefor sigset_t for mips >> >> >> >> >> >> On Mon, Nov 9, 2009 at 2:12 PM, Henry Yei <[email protected]> >> wrote: >> >> >> >> -----Original Message----- >> >> >> >> From: Subrata Modak [mailto:[email protected]] >> >> >> >> Sent: Monday, November 09, 2009 9:40 AM >> >> >> >> To: Henry Yei >> >> >> >> Cc: Garrett Cooper; LTP Mailing List >> >> >> >> Subject: RE: [LTP] [PATCH][ppoll01] fix testcase 1, 5: pass >> >> correct >> >> >> >> valuefor sigset_t for mips >> >> >> >> >> >> >> >> On Mon, 2009-11-02 at 14:09 -0800, Henry Yei wrote: >> >> >> >> > Subrata, >> >> >> >> > >> >> >> >> > I've included Nicolas Joly's suggested removal of progdir to >> >> force >> >> >> >> the test to create all test files in the temp directory in the >> >> >> attached >> >> >> >> patch. >> >> >> >> > >> >> >> >> > I'm not sure why tst_exit() is needed when calling >> cleanup(), >> >> as >> >> >> it >> >> >> >> would be called by cleanup(). I can add it back, if needed. >> Any >> >> >> other >> >> >> >> comments on the the appropriateness of "sizeof(sigmask)" would >> >> also >> >> >> be >> >> >> >> welcomed. There are a few other tests with similar issues. >> >> >> >> > >> >> >> >> >> >> >> >> Henry, >> >> >> >> >> >> >> >> Probably because of heavy changes recently, this patch is not >> >> >> applying >> >> >> >> properly. Can you please rebase your changes: >> >> >> >> >> >> >> >> patching file testcases/kernel/syscalls/ppoll/ppoll01.c >> >> >> >> Hunk #3 FAILED at 98. >> >> >> >> Hunk #4 FAILED at 127. >> >> >> >> Hunk #5 FAILED at 185. >> >> >> >> Hunk #6 FAILED at 252. >> >> >> >> 4 out of 6 hunks FAILED -- saving rejects to file >> >> >> >> testcases/kernel/syscalls/ppoll/ppoll01.c.rej >> >> >> >> >> >> >> >> Regards-- >> >> >> >> Subrata >> >> >> >> >> >> >> >> > > -----Original Message----- >> >> >> >> > > From: Subrata Modak [mailto:[email protected]] >> >> >> >> > > Sent: Monday, November 02, 2009 5:59 AM >> >> >> >> > > To: Henry Yei >> >> >> >> > > Cc: Garrett Cooper; LTP Mailing List >> >> >> >> > > Subject: Re: [LTP] [PATCH][ppoll01] fix testcase 1, 5: >> pass >> >> >> correct >> >> >> >> > > valuefor sigset_t for mips >> >> >> >> > > >> >> >> >> > > Any further work on this patch ? >> >> >> >> > > >> >> >> >> > > Regards-- >> >> >> >> > > Subrata >> >> >> >> > > >> >> >> >> > > On Wed, 2009-10-28 at 12:08 -0700, Henry Yei wrote: >> >> >> >> > > > > -----Original Message----- >> >> >> >> > > > > From: Garrett Cooper [mailto:[email protected]] >> >> >> >> > > > > Sent: Tuesday, October 27, 2009 7:57 PM >> >> >> >> > > > > To: Henry Yei >> >> >> >> > > > > Cc: LTP Mailing List >> >> >> >> > > > > Subject: Re: [LTP] [PATCH][ppoll01] fix testcase 1, 5: >> >> pass >> >> >> >> correct >> >> >> >> > > > > value for sigset_t for mips >> >> >> >> > > > > >> >> >> >> > > > > On Tue, Oct 27, 2009 at 6:32 PM, Henry Yei >> >> <[email protected]> >> >> >> >> wrote: >> >> >> >> > > > > > All, >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > The patch attached for ppoll01: >> >> >> >> > > > > > >> >> >> >> > > > > > cleans up spacing and code style >> >> >> >> > > > > > >> >> >> >> > > > > > removes unneeded/broken debug option parsing >> >> >> >> > > > > > >> >> >> >> > > > > > if test fails, it prints out the last failed errno, >> >> rather >> >> >> >> than >> >> >> >> > > just >> >> >> >> > > > > the >> >> >> >> > > > > > last errno(which may have been successful) >> >> >> >> > > > > > >> >> >> >> > > > > > passes in correct size of sigset_t for mips >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > Signed-off by: Henry Yei <[email protected]> >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > >> >> >> >> > > > > > This patch was tested/passed on x86, mips(little >> >> endian), >> >> >> and >> >> >> >> > > > > ppc_82xx . >> >> >> >> > > > > >> >> >> >> > > > > 1. Please keep tst_exit() at the bottom. >> >> >> >> > > > > 2. Why can't we do sizeof(sigsetmask) ? >> >> >> >> > > > > >> >> >> >> > > > > Thanks, >> >> >> >> > > > > -Garrett >> >> >> >> > > > >> >> >> >> > > > Garrett, >> >> >> >> > > > >> >> >> >> > > > You want me to put tst_exit() at the bottom, even though >> it >> >> is >> >> >> >> called >> >> >> >> > > within cleanup()? >> >> >> >> > > > You mean sizeof(sigmask) ~ sizeof(sigset_t), right? That >> >> code >> >> >> was >> >> >> >> > > commented out since version 1.1 which I thought was odd >> too, >> >> so >> >> >> I >> >> >> >> did >> >> >> >> > > check sizeof(sigmask). >> >> >> >> > > > However, sizeof(sigmask) on mips returns 128 which does >> not >> >> >> seem >> >> >> >> > > correct. I haven't checked the other architectures, but I >> >> >> suspect >> >> >> >> some >> >> >> >> > > might have a similar mismatch. Apparently sigset_t may >> have >> >> >> >> different >> >> >> >> > > values when returned from glibc vs kernel space. More >> >> >> information >> >> >> >> from >> >> >> >> > > this post: >> >> >> >> > > > >> >> >> >> > > > http://lists.uclibc.org/pipermail/uclibc/2009- >> >> >> January/041850.html >> >> >> > >> >> >> > >> >> >> > Subrata, >> >> >> > >> >> >> > I've included the ppoll01 patch which should apply cleanly to >> the >> >> >> latest in CVS. >> >> >> > >> >> >> >> >> >> Hi Henry, >> >> >> Could you change this line: >> >> >> >> >> >> + tst_resm(TFAIL, "%s failed - errno >> = >> >> %d : %s", >> >> >> TCID, >> >> >> + last_failed_errno, >> >> >> strerror(last_failed_errno)); >> >> >> >> >> >> to this please? >> >> >> >> >> >> + tst_resm(TFAIL | TTERRNO, "%s >> failed", >> >> TCID); >> >> >> >> >> >> last_failed_errno should become TEST_ERRNO if this change is >> >> done, >> >> >> IIRC. >> >> >> >> >> > >> >> > I've attached the ppoll01 patch with Garrett's suggested changes >> to >> >> use TEST_ERRNO. >> >> >> >> You could just completely zap last_failed_errno and make it >> test_errno >> >> instead. That's what I was suggesting with the previous reply. >> >> On Wed, Nov 11, 2009 at 11:25 AM, Henry Yei <[email protected]> wrote: >> > Wouldn't TEST_ERRNO be the errno of the last TEST() that was run, >> rather than the last failed one? >> > >> > >> > >> Yup! The goal is to drop the additional variable >> (last_failed_errno) so that only TEST_ERRNO is required and then all >> you need to do is use TTERRNO and you're golden :). >> Thanks, >> -Garrett > > Attached is the poll01 patch with last_failed_errno and associated code > removed.
AHA! I figured out why sigset is sized differently on mips32 for you, compared to other architectures. Just like there's a compat_caddr_t datatype (which is explicitly set to char *), there's also a compat_sigset_t as well. Let me see if I can figure out this mystery, but my guess is that you're using n32 based mips, most likely on an Octeon model ;) ;)... which would account for what you're seeing here and possibly account for why the test consistently hangs on our Cavium boards at work (our version of Linux has been hacked so bad though to take shortcuts in the design process instead of following proper design procedure that it makes me cry inside...). Thanks, -Garrett ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Ltp-list mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ltp-list
