Re: [arch-dev-public] [arch-dev] prepared statements in recent mysql update broken?
On Thursday 11 June 2009 01:38:59 Pierre Schmitz wrote:
Hi,
after upgrading mysql from 5.1.34 to.35 I noticed some quite strange behaviour. I am not done with my tests yet but it seams that at least with PHP prepared statements are broken under some conditions.
E.g. The example from http://de2.php.net/manual/en/mysqli-stmt.execute.php gave me the output:
Stuttgart (DEU,0) Bordeaux (FRA,0)
It should be
Stuttgart (DEU,Baden-Wuerttemberg) Bordeaux (FRA,Aquitaine)
So somehow the binding of the third variable fails. Downgrading mysql to .34 solves the problem. Recompiling php (maybe there was a hidden ABI change like there was some in the past) did not help either.
Me again. So, I tried to rebuild .34. The result was again a broken mysql. So its eems its not mysql itself but one of its static deps or gcc. It seems the old package was build with gcc 4.3. -- Pierre Schmitz, http://users.archlinux.de/~pierre
On Friday 19 June 2009 06:41:08 Pierre Schmitz wrote:
Me again. So, I tried to rebuild .34. The result was again a broken mysql. So its eems its not mysql itself but one of its static deps or gcc. It seems the old package was build with gcc 4.3.
I tried rebuilding with old gcc, gcc-libs, glibc and bintuils; no luck. I am running out of options... -- Pierre Schmitz, http://users.archlinux.de/~pierre
On Fri, 2009-06-19 at 13:03 +0200, Pierre Schmitz wrote:
On Friday 19 June 2009 06:41:08 Pierre Schmitz wrote:
Me again. So, I tried to rebuild .34. The result was again a broken mysql. So its eems its not mysql itself but one of its static deps or gcc. It seems the old package was build with gcc 4.3.
I tried rebuilding with old gcc, gcc-libs, glibc and bintuils; no luck. I am running out of options...
Are you sure it's mysql, and not libmysqlclient that breaks this?
On Friday 19 June 2009 13:10:45 Jan de Groot wrote:
Are you sure it's mysql, and not libmysqlclient that breaks this?
No idea. In my tests I allways rebuilt all three mysql packages. (libs, mysqld and clients) And its not related to php only; see for example: http://bbs.archlinux.org/viewtopic.php?id=74331 http://bbs.archlinux.org/viewtopic.php?id=74295 http://bbs.archlinux.org/viewtopic.php?id=74224 Allways binding variables fails or even worth you get wrong answers or you write wrong data into the db. -- Pierre Schmitz, http://users.archlinux.de/~pierre
Pierre Schmitz wrote:
On Thursday 11 June 2009 01:38:59 Pierre Schmitz wrote:
Hi,
after upgrading mysql from 5.1.34 to.35 I noticed some quite strange behaviour. I am not done with my tests yet but it seams that at least with PHP prepared statements are broken under some conditions.
E.g. The example from http://de2.php.net/manual/en/mysqli-stmt.execute.php gave me the output:
Stuttgart (DEU,0) Bordeaux (FRA,0)
It should be
Stuttgart (DEU,Baden-Wuerttemberg) Bordeaux (FRA,Aquitaine)
So somehow the binding of the third variable fails. Downgrading mysql to .34 solves the problem. Recompiling php (maybe there was a hidden ABI change like there was some in the past) did not help either.
Me again. So, I tried to rebuild .34. The result was again a broken mysql. So its eems its not mysql itself but one of its static deps or gcc. It seems the old package was build with gcc 4.3.
Just a random though... is there any file differences between the original .34 and a rebuilt one? Also, is there a bug report open to track this? I did a quick search and did not find one. Allan
On Tuesday 23 June 2009 17:26:29 Allan McRae wrote:
Just a random though... is there any file differences between the original .34 and a rebuilt one?
Also, is there a bug report open to track this? I did a quick search and did not find one.
If I have some time I'll open a task in our tracker today. So far: Only libmysqlclient is affected; the other two can be updated without problems. There is no difference in the filelist between .34 and .35. -- Pierre Schmitz, http://users.archlinux.de/~pierre
Pierre Schmitz wrote:
On Tuesday 23 June 2009 17:26:29 Allan McRae wrote:
Just a random though... is there any file differences between the original .34 and a rebuilt one?
Also, is there a bug report open to track this? I did a quick search and did not find one.
If I have some time I'll open a task in our tracker today. So far: Only libmysqlclient is affected; the other two can be updated without problems.
There is no difference in the filelist between .34 and .35.
So, if it is only libmysqlclient, should we push the rebuild for mysql so we can move the bash/readlines stuff. It gets no more broken than what is currently in extra... Allan
On Wednesday 24 June 2009 04:20:19 Allan McRae wrote:
So, if it is only libmysqlclient, should we push the rebuild for mysql so we can move the bash/readlines stuff. It gets no more broken than what is currently in extra...
It's hard to decide on which level of broken we are atm. ;-) I just opened a new task: http://bugs.archlinux.org/task/15242 Anyway: its not usefull to block the move of readline and bahs because of this. People using mysql shouldn't update and downgrade to the old libmysqlclient-5.1.34 -- Pierre Schmitz, http://users.archlinux.de/~pierre
participants (3)
-
Allan McRae
-
Jan de Groot
-
Pierre Schmitz