Ensure handle is valid and pm_errno is reset when calling into API

We didn't do due diligence before and ensure prior pm_errno values
weren't influencing what happened in further ALPM calls. I observed one
case of early setup code setting pm_errno to PM_ERR_WRONG_ARGS and that
flag persisting the entire time we were calling library code.

Add a new CHECK_HANDLE() macro that does two things: 1) ensures the
handle variable passed to it is non-NULL and 2) clears any existing
pm_errno flag set on the handle. This macro can replace many places we
used the ASSERT(handle != NULL, ...) pattern before.

Several other other places only need a simple 'set to zero' of the
pm_errno field.

Signed-off-by: Dan McGee <dan@archlinux.org>

1 files changed, 4 insertions, 3 deletions

diff --git a/lib/libalpm/sync.c b/lib/libalpm/sync.c
index d65f5e20..05735730 100644
--- a/lib/libalpm/sync.c
+++ b/lib/libalpm/sync.c
@@ -55,11 +55,12 @@
  */
 pmpkg_t SYMEXPORT *alpm_sync_newversion(pmpkg_t *pkg, alpm_list_t *dbs_sync)
 {
-	ASSERT(pkg != NULL, return NULL);
-
 	alpm_list_t *i;
 	pmpkg_t *spkg = NULL;
 
+	ASSERT(pkg != NULL, return NULL);
+	pkg->handle->pm_errno = 0;
+
 	for(i = dbs_sync; !spkg && i; i = i->next) {
 		spkg = _alpm_db_get_pkgfromcache(i->data, alpm_pkg_get_name(pkg));
 	}
@@ -89,7 +90,7 @@ int SYMEXPORT alpm_sync_sysupgrade(pmhandle_t *handle, int enable_downgrade)
 	pmdb_t *db_local;
 	alpm_list_t *dbs_sync;
 
-	ASSERT(handle != NULL, return -1);
+	CHECK_HANDLE(handle, return -1);
 	trans = handle->trans;
 	db_local = handle->db_local;
 	dbs_sync = handle->dbs_sync;