[PATCH] vformat to compile on Windows

[henrik]

The vformat/tests/support.c file does not compile on windows (MinGW).

Patch attached.

Any comments?

[dgollub]

why not using g_mkstemp_full() directly from glib?

[henrik]

Because it creates a temporary FILE and what we need is a DIRECTORY.

[dgollub]

Try this instead:

char *setup_testbed(const char *fkt_name)
{
#ifndef _WIN32  
        setuid(65534);
#endif
        char *testbed = g_strdup_printf("%s/testbed.XXXXXX", g_get_tmp_dir());
        char *command = NULL, *dirname = NULL;
#ifdef _WIN32
        if(g_file_test(testbed, G_FILE_TEST_IS_DIR))
                destroy_testbed(g_strdup(testbed));

        if(g_mkdir(testbed,0777) < 0){
                        abort();
        }
#else /* WIN32 */
        if (!mkdtemp(testbed))
                abort();
#endif /* WIN32 */

[...]

this is code from OpenSync? trunk: trunk/tests/support.c

Simliar/redudant code ... we should kill this redudant code in near future.

[henrik]

As far as I can see, this approach would - on WIN32 - use a fixed name for the testbed (testbed.XXXXXX), not a pseudo-random name.

This is bad if you run test-cases in parallel - even if it is for different components (opensync, vformat, ...).

And although the temp directory on WIN32 is normally not shared between users, I don't think there is anything preventing it from being shared between users, which would then also be a problem.

[dgollub]

This code is only for unit-testing. Currently i see no reason to run tests in parallel ... This kind of code is only run by continous build hosts or developers - not regular users.

Feel free to replace XXXXX with a random number or so - or lock the directory for windows.

I don't want to introduce too much code from different projects - since we would have to maintain this as well. And we defenitly should try to avoid duplicate code in OpenSync? and vformat.

Since nobody on windows complained about windows temp direcory XXXXX for OpensYnc? testsuite - i doubt someone will for vformat ...