> On Tuesday 28 July 2009 10:18:19 [email protected] wrote: >> While running the tests I encountered something strange. The test for >> pdf_fsys_get_temp_file_name() failed because the function returned NULL >> (pdf_fsys_file_open_005 failed for the same reason). When I set the >> CK_FORK environment varaible to "no" in order to debug this, all test >> ran >> fine?! > I've seen problems that change depending on CK_FORK. > > Lets look at this. When you run "normally", the tests get run as a forked > process. This means that they are not impacted by other tests. That isn't > the > case with CK_FORK="no". > > Perhaps what is happening is that some other test(s) are performing > initialisation that helps your test pass? > > Brad
Hi, you're right. pdf_fsys_disk_get_temp_path_name() calls pdf_text_new_from_unicode(), which needs the text module to be initialized by pdf_text_init(). Some other test running before the failing one called pdf_init_text() so everything worked with CK_FORK=no. I'll adjust my tests accordingly. best regards, Hans Hohenfeld
