From: Mark Dickinson Date: Sun, 13 Dec 2009 20:03:21 +0000 (+0000) Subject: Make sure that test_capsule always returns NULL on error; this may X-Git-Tag: v3.2a1~2037 X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=6d138f1f450aa1ef0f95b55c006244d7c65c3c48;p=python Make sure that test_capsule always returns NULL on error; this may help diagnose the sporadic test_capi failures on Solaris. --- diff --git a/Modules/_testcapimodule.c b/Modules/_testcapimodule.c index f41d4fec03..aca860adc5 100644 --- a/Modules/_testcapimodule.c +++ b/Modules/_testcapimodule.c @@ -1298,6 +1298,21 @@ test_capsule(PyObject *self, PyObject *args) if (error) { return raiseTestError("test_capsule", error); } + /* 13/12/2009: something is causing test_capi to fail occasionally on + the Solaris buildbot, with the output: + + internal test_L_code + internal test_Z_code + internal test_bug_7414 + internal test_capsule + XXX undetected error + internaltest test_capi crashed -- : No module named datetime + + It seems possible that test_capsule is raising an exception but + failing to return NULL. Do a PyErr_Occurred check to find out. + */ + if (PyErr_Occurred()) + return NULL; Py_RETURN_NONE; #undef FAIL }