From: Garrett C. <yan...@gm...> - 2010-07-23 22:03:53
|
On Jul 23, 2010, at 11:48 AM, Nicolas Joly wrote: > On Fri, Jul 23, 2010 at 10:56:58AM -0700, Garrett Cooper wrote: >> On Fri, Jul 23, 2010 at 10:17 AM, Nicolas Joly <nj...@pa...> wrote: >>> >>> Hi, >>> >>> The attached patch do some cleanup in the abort01 testcase ... >>> >>> 1) The attempt to remove the generated core is wrong. The hard-coded >>> `core' name is bad and the test temporary directory removal already >>> take care of this. >>> >>> 2) Make the test fail gracefully if the running environment does not >>> allow generating core files. >>> >>> njoly@lanfeust [syscalls/abort]> ./abort01 >>> abort01 1 TPASS : Test passed >>> njoly@lanfeust [syscalls/abort]> (ulimit -c 0 && ./abort01) >>> abort01 1 TCONF : core file size limit must be greater than 0. >>> >>> Signed-off-by: Nicolas Joly <nj...@pa...> >> >> Looks ok, but is there any particular reason why you removed the unlink(2) call? > > The `unlink("core")' one ? > > Well, at least RHEL do use `proc.<pid>' when generating core file > names ... Not speaking about admins that might have configured it to > use another scheme. > > IMO, relying on a specific name, for a configurable system, is a bad > idea; and can only lead to problems. > > By example, this test from mkdir09.c cannot succeed on most RHEL > systems: > > /* Check for core file in test directory. */ > if (access("core", 0) == 0) { > tst_resm(TWARN, "\tCore file found in test directory."); > tst_exit(); > } Yeah. Perhaps it might be a good idea to integrate a tool into LTP which detects corefile names, like what's described here: http://aplawrence.com/Linux/limit_core_files.html . Cheers, -Garrett |