<div class="gmail_quote">On Mon, Jun 21, 2010 at 1:18 PM, Convey, Christian J CIV NUWC NWPT <span dir="ltr"><<a href="mailto:christian.convey@navy.mil">christian.convey@navy.mil</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
Sorry, that was embarrassing. Pretend I had said, "CTestTestFile.cmake" instead of "CTestConfig.cmake".<br></blockquote><div><br>Well, in that case.... CMake writes the CTestTestFile.cmake files whenever it configures a build tree. (So, from a ctest -S script, during the "ctest_configure(...)" call.) The ctest executable will read the CTestTestFile.cmake files when it needs to execute tests: when run from the command line to execute tests, or when processing a ctest_test(...) command in a -S script...<br>
<br>So again: only one writer, at one well-defined time. I don't think there should be any problems with respect to the CTestTestFile.cmake files. The confusion arises perhaps because ctest is used as a "simple" test driver and also as a full configure/build/test/run-a-dashboard driver. The CTestTestFile.cmake files are used when it is acting as a simple test driver. You would write a script to be processed with -S, or run a -D Nightly or -D Experimental dashboard separately: that is not connected to the CTestTestFiles at all.<br>
<br>I suppose you could probably contrive a case where an executing test actually re-configures the build tree that it is in, but I'd be surprised if that happens in practice. If shown to be problematic, and reported with reasonable steps to reproduce to the CMake bug tracker, we could add code to detect such pathological cases and fail with an error...<br>
<br><br>HTH,<br>David<br><br></div></div><br>