The output file used for memory checker runs must be unique for every test run
in parallel, so simply make them unique for every test run. Simply use the test
index to avoid collisions.
This reverts commit 6187876dea.
It was actually possible before to have paths with spaces in them, the spaces
just need to be quoted. This way spaces will work as argument separators.
Checkout [1] as an example of the test failures. In that particular cases, the
failures is caused by the plus sign in the path being pass unescaped
(buildd-cmake_2.8.9-1~bpo60+1-armel-3Lvkef) to the regexp.
In addition to failures in the log, the following new tests also fail in 2.8.11:
243 - CTestTestMemcheckUnknown (Failed)
244 - CTestTestMemcheckUnknownQuoted (Failed)
248 - CTestTestMemcheckDummyValgrindFailPre (Failed)
249 - CTestTestMemcheckDummyValgrindFailPost (Failed)
250 - CTestTestMemcheckDummyPurify (Failed)
251 - CTestTestMemcheckDummyBC (Failed)
253 - CMake.List (Failed)
[1] https://buildd.debian.org/status/fetch.php?pkg=cmake&arch=armel&ver=2.8.9-1~bpo60%2B1&stamp=1369243896
Both <string> and <cmSystemTools.h> include <stdio.h> one or the other way. It
looks like the CMake header adds a define that will cause another part of the
system header to be used, resulting in conflicting definitions.
All supported memory checkers now write their output to a file. Use a dummy
checker that ignores the given filename and ensure that the missing file is
reported as error.
The code for the tests is basically the same for all those subtests, so have
one template and configure that for as many tests as possible to make it
easier maintainable.
The memory checker command can't be quoted at this point, because previously it
has been tested that the given file exists, which will fail if the name is
quoted. The CTestTestMemcheckUnknown test aimed to test this case, has always
failed to do so and serves no useful purpose therefore.
The dummy memory tester implementation now understands the command line
switches for all memory checkers to redirect the output to a file. This avoids
triggering the error cases for BoundsChecker and Purify because the output file
does not exist.