[EMAIL PROTECTED] wrote:
Jon, I've cc'd my reply to dbi-dev list (which you won't be able to post
to unless you are a subscriber) - I hope you don't mind.
Martin,
After another Google search, I may have found your patches…
_http://osdir.com/ml/lang.perl.modules.dbi.sybase.devel/2006-10/msg00048.html_
Links to:
_http://osdir.com/ml/lang.perl.modules.dbi.sybase.devel/2006-10/binnya3doknmF.bin_
Do you have a sense that the trouble is mostly with the test suite, or
are the patches something I should lobby for in approving changes to
DBD::ODBC 1.13 on all platforms? (Solaris, AIX, etc)
Fix bug in dbd_bind_ph which leads to the error Can't change param 1
maxlen (51->50) after first bind in the 20SqlServer test. This is caused
by svGROW in _Perl_
<http://osdir.com/ml/lang.perl.modules.dbi.sybase.devel/2006-10/msg00048.html>
5.8.8 being changed to possibly grow by more than you asked (e.g. up to
the next longword boundary).
I'm wondering if I should perhaps create a sub-set of your patch to
address the specific issue for testing, or if all your fixes are in
general use…
Thank you very much!!
The specific problem you refer to is not in the test suite but in a
change to svGROW in perl in 5.8.8. svGROW used to return the exact
amount of memory you asked for and DBD::ODBC relied on that and now it
can return more than you asked for which breaks that test.
As for comments on the changes:
Fix bug in dbd_bind_ph which leads to the error
Can't change param 1 maxlen (51->50) after first bind
in the 20SqlServer test. This is caused by svGROW in Perl 5.8.8 being
changed to possibly grow by more than you asked (e.g. up to the next
longword boundary).
This needs to be in DBD::ODBC - no doubt about it.
Fix problem with binding undef as an output parameter. Reported by
Stephen More with IBM's ODBC driver for iSeries.
This needs to be in DBD::ODBC. Stephen verified the fix at the time.
Removed comment delimiters in comments in dbdimp.h leading to warnings.
These are harmless but confuse people who see unnecessary warnings.
Removed some unused variable declarations leading to warnings.
As above.
Fixed use of unitialised variable in dbd_bind_ph where an undef is bound
and tracing is enabled.
straight forward bug - needs to be in DBD::ODBC.
and changes:
Removed PerlIO_flush calls as it is believed they are not required.
See comment from Tim Bunce about this - PerlIO_flush calls are probably
not required any more so this should be in DBD::ODBC.
Add logging for whether SQLDescribeParam is supported.
Not necessary but useful.
Also note, that the current test suite fails some tests when connected
to MS SQL Server 2005 or Express - I don't have fixes for those as yet
but I will be working on it.
I have since come up with changes for this to make it work but I don't
have them to hand right now. I may have already posted them to dbi-users.
Jeff Urlwin, who looks after DBD::ODBC usually applies the patches I've
submitted in the past but I have not heard from him in a while.
I have used a version of DBD::ODBC with these changes for ages now.
Martin
--
Martin J. Evans
Easysoft Limited
http://www.easysoft.com
