[pacman-dev] [PATCH] lib/util: call _alpm_log before setting handle->pm_errno

[Dave Reisner]

This is an unfortunate chain of events. RET_ERR and RET_ERR_VOID will
eventually call CHECK_HANDLE, which resets the handle's pm_errno member.
Dan probably had a reason for doing this, so we merely switch the order
of operations in the RET_ERR macros to avoid stomping on our pm_errno.

Signed-off-by: Dave Reisner <d at falconindy.com>
---
Awesome bug. Here's what happens when the DB is locked:

$ sudo pacman -Sy
:: Synchronizing package databases...
error: failed to init transaction (unexpected error)

I'm still not sure why CHECK_HANDLE will always reset handle->pm_errno. Seems
like it's some sort of lazy initialization that should be getting done
elsewhere. There may be a better way to solve this, if the reset isn't actually
necessary.

 lib/libalpm/util.h |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/lib/libalpm/util.h b/lib/libalpm/util.h
index 0549c81..778e20f 100644
--- a/lib/libalpm/util.h
+++ b/lib/libalpm/util.h
@@ -61,12 +61,14 @@
 
 #define ASSERT(cond, action) do { if(!(cond)) { action; } } while(0)
 
-#define RET_ERR_VOID(handle, err) do { (handle)->pm_errno = (err); \
+#define RET_ERR_VOID(handle, err) do { \
 	_alpm_log(handle, PM_LOG_DEBUG, "returning error %d from %s : %s\n", err, __func__, alpm_strerror(err)); \
+	(handle)->pm_errno = (err); \
 	return; } while(0)
 
-#define RET_ERR(handle, err, ret) do { (handle)->pm_errno = (err); \
+#define RET_ERR(handle, err, ret) do { \
 	_alpm_log(handle, PM_LOG_DEBUG, "returning error %d from %s : %s\n", err, __func__, alpm_strerror(err)); \
+	(handle)->pm_errno = (err); \
 	return (ret); } while(0)
 
 #define DOUBLE_EQ(x, y) (fabs((x) - (y)) < DBL_EPSILON)
-- 
1.7.5.4