free(virtTestLogContentAndReset()); cmd = virCommandNew(abs_builddir "/commandhelper-doesnotexist"); - if (virCommandRun(cmd, NULL) == 0) + if (!cmd || virCommandRun(cmd, NULL) == 0)
The API explicitly does *not* require you to check !cmd after virCommandNew. virCommandRun() and other APis will check that for you and return ENOMEM.
That was my suspicion too but then I looked at virCommandRun implementation and saw this: if (!cmd || cmd->has_error == -1) { virCommandError(VIR_ERR_INTERNAL_ERROR, "%s", _("invalid use of command API")); return -1; } if (cmd->has_error == ENOMEM) { virReportOOMError(); return -1; } That is, if virCommandNew() returns NULL for either reason, by running virCommandRun() on it, the original error (probably OOM) gets overwritten by an internal error. Jirka