Change Log for DBD::Informix for DBI
- Written by Alligator Descartes <descarte@arcana.co.uk>
- Rewritten by Jonathan Leffler <j.leffler@acm.org>
0.59 Release
------------
98-03-11: Make a release to try and resolve ongoing problems at AirTouch.
98-03-04: Chuck Collins <Chuck.Collins@AirTouch.COM> discovered that the
dbd_ix_preparse() routine gets confused by DATETIME and INTERVAL
literals -- it was treating '30:20:10' as '30??' because the
literals are not inside a string. Fixed by not recognizing
Oracle-style numeric place-holders (Oracle-style named place-holders
were already ignored because they can be confused with dbase:table).
98-02-20: Chuck Collins <Chuck.Collins@AirTouch.COM> discovered an assertion
failing in MODE ANSI databases when an UPDATE didn't update any
rows. Fix assertion to cover this case. Testing reveals that if
you create a database, the logging status, etc, are not set
correctly. This also causes an assertion to fire.
98-02-17: DBI 0.93 released with fix for '.xsi' problem.
98-02-11: Problem with Informix.xsi resolved. Needs extra dependency in
generated Makefile, which is triggered by the DBI version.
98-02-06: Document problems with DBI 0.92 (no Informix.xsi file).
0.58 Release
------------
98-01-15: Finish up NT build work, and release in time for the Win32 Perl
Resource Book deadline (phew) and also just before switching jobs
from Informix to Visa. Fix Makefile.PL and README file to make
warnings about using ld to create shared libraries with gcc more
stringent. Kevin White <kwhite@redroof.com> and Michael Chandler
<mpchandler@fedex.com> both had major problems on HP-UX 10.20
with this, and the problems were not fully resolved at release
time. One issue was that the notation 'ENVVAR=value command' did
not seem to be setting the environment variable, but Makefile.PL
assumes this will work. Maybe it should use Perl's %ENV instead.
98-01-14: More work on NT builds with help from Ken Scott <ken@apisys.net>
and Patrick-James Dionisio <dionisio@netscape.com>.
98-01-13: Use warn instead of die to report lower than ideal DBI version in
Makefile.PL. Also make corrections for NT builds based on
feedback from Ken Scott <kscott@pcisys.net>.
98-01-12: Fix dbd_ix_preparse() so it does not convert :alpha into ? after
Chuck Collins <Chuck.Collins@AirTouch.COM> pointed out that
SELECT * FROM sysmaster:systabnames gives a syntax error in 0.57.
98-01-07: Initial attempt at integrating Makefile.NT into Makefile.PL.
98-01-06: Modify Makefile.PL to give a reason for failure when stating that
you did not read the README file after Michael Chandler
<mpchandler@fedex.com> ran into problems identifying the source
of such a failure.
98-01-05: Update copyright dates. Change email address for maintainer.
97-12-03: Simplify the version testing to ignore the lists of libraries.
The test compilation ensures that everything is OK, or not.
Also, simply use the local copy of odbctype.h, without checking
whether any ODBC is available or not.
97-11-18: Steven Hronek <hronek@lsil.com> asked whether ESQL/C was still
necessary. Ensure that README is absolutely clear that it is,
and mention DBD::ODBC as a possible alternative.
0.57 Release
------------
97-11-17: Final clean up and testing. Version 0.56 released.
97-10-19: Handle the 'esql -libs' process rather differently, because of
the bug in older versions of the esql script which do not echo
the name of the netstub library when necessary (a problem found
by Rodger Anderson <rodger@hpbs2245.boi.hp.com> in July, but only
now resolved). Record information about compiling with c4gl only
in Informix.Licence, after answering questions from Paul Falbe
<paul@cassens.com>.
97-10-08: Change the name of the sqlca element of imp_dbh to ix_sqlca to
avoid the macro renaming used in threaded versions of ESQL/C such
as on Windows 95 and Windows NT. This change and success with a
Windows 95 and Windows NT port reported by Harald Ums
<Harald.Ums@sevensys.de>. Also a change to esqlc.h, triggered by
the need to use sqlproto.h, and use io.h in place of unistd.h in
ixblob.ec, and do not release 'unreleased' blobs... Makefile.PL
still to be upgraded.
97-10-08: Record Informix ESQL/C's ability to change the environment
without Perl knowing about it, after the problem found by Jay
Ludoviconi <jludovic@paging.mot.com> was tracked down.
97-10-03: Add Working.Versions file after a suggestion by Nuno Carneiro de
Moura <ncmoura@net.mailcom.pt>
97-08-13: Add file hints/static.build with info about building statically
linked Perl executables from Kent S Gordon <kgor@inetspace.com>.
Clean up function entry and exit reporting.
97-08-01: Add _tables and _columns metadata methods.
97-07-29: Update to use the templates from DBI v0.89, etc. Amend README
after Kent S Gordon <kgor@inetspace.com> noted that the engine
version information could be useful.
97-07-21: Add dbd_ix_preparse() (based on code in DBD::ODBC 0.15) to
pre-process an SQL statement, replacing Oracle-style :x notation
with question mark place holders. This allows DBD::Informix to
support $sth->{NUM_PARAMS} properly, and reinstates functionality
present in DBD::Informix 0.24 but summarily removed in version
0.25. Interestingly, of course, noone has complained, or even
commented on this.
97-07-16: Kenji Hino <kenji@alcatraz.net> found that the test t/blob02.t
failed for logged OnLine databases because of the changed
AutoCommit handling; alter the criterion for doing $dbh->commit.
Jason Bodnar <jcbodnar@mail.utexas.edu> found that the
documentation for $dbh->do() was faulty; add the undef parameter
for the unused \%attr documented by DBI.
97-07-14: Deal with the last of the deprecated attributes. Work around the
oddities (changes) introduced by ESQL/C 9.12.
0.56 Release
------------
97-07-10: Final clean up and testing (and workaround for problem in DBI
0.85 when connections fail at connect time).
Version 0.56 released.
97-07-09: Establish that DBI->connect(...,{AutoCommit=>0}) does not set the
AutoCommit attribute until after the basic connect operation is
complete. DBD::Informix therefore cannot reject connections on
unlogged databases in the basic connect operation. Modified
Informix.Licence after Raj Goel <frostbit@brainlink.com> needed
to ask questions. The difficulty with 'newbies' is that they
quickly cease to be newbies so you need a continuous supply of
new newbies. Fix documentation in Informix.pm for AutoCommit,
and behaviour of ix_InTransaction attribute. Many thanks to Hank
Robinson <cudau@csv.warwick.ac.uk> for his help clarifying the
required behaviour of AutoCommit, and apologies for omitting his
name in the original version of the 0.56 release notes.
97-07-08: Make $dbh->{ix_AutoErrorReport} a synonym for $h->{PrintError}.
97-07-07: Jason Bodnar <jcbodnar@mail.utexas.edu> pointed out that
$sth->execute() is not returning undef when an error occurs.
97-06-25: Roderich Schupp <roderich@nodebonn.muc.bmw.de> found a similar
problem using $dbh->do($stmt, undef, $arg1, $arg2), core dumping
in the link code. This could be reproduced with the as-released
0.55 DBD::Informix but appears to be fixed, presumably by the fix
for Fred's problem. Also, ensure the t/dbase.t unsets both
$ENV{DBI_DSN} and $ENV{DBI_DBNAME} to allow the implicit connect
test to work (deduced from test output sent by Fred Arnold).
Move comments about Solaris compiler into new file hints/solaris.
Move README.M88K.SVR4 to hints/M88K.SVR4.
97-06-24: Found problems in DBI 0.84 where PrintError, RaiseError, ChopBlanks
as handled by DBI cannot be unset -- Tim confirmed it is a bug in
DBI and will be fixed in the next release.
97-06-23: Core dump problem found by Fred Arnold <ftarnol@uswest.com> in the
code in link.c and the way it is used. One part of the problem
was that the IMPSET flag for a statement was not being set soon
enough, so that DBI was not releasing the statement from the
linked lists correctly. The temporary fix includes a bodge that
converts the deleted link into a single item linked list, so that
when it is released again, it doesn't screw things up. This
needs to be cleaned up. A 0.56a1 release was sent to Fred to get
him over the immediate problem.
97-06-17: Jason Bodnar <jcbodnar@mail.utexas.edu> pointed out that BEGIN
WORK is not being recognized via $dbh->do(). Testing showed a
problem during disconnect on 7.1x servers (related to the B42204
problem fixed on 97-05-30). Fixing the disconnect was relatively
simple. Fixing the BEGIN WORK issues required the removal of the
custom implementation of $dbh->do(), and some other changes.
97-06-13: Where there's one bug, there are usually others. The other array
attributes for statements (such as $sth->{NAME}) were not being
mortalized and could be made to eat up memory if used inside an
inner loop. Upgrade non-distributed test bug001.t to catch these
leaks too.
97-06-12: Fix trailing blanks on $sth->{NAME} attributes after problem
reported by Roderich Schupp <roderich@nodebonn.muc.bmw.de>. Also
fix t/chopblanks.t not to create VARCHAR columns in SE.
97-05-30: Fix memory leak when using the $sth->{ix_sqlerrd} and
$sth->{ix_sqlwarn} attributes repeatedly, reported by Oyvind
Gjerstad <ogj@tglobe1.tollpost.no>. The arrays in newSqlerrd()
and newSqlwarn() were not being mortalized as they should be.
97-05-30: Only set -Ae on HP-UX when using cc, as reported by Nathan
Neulinger <nneul@umr.edu> who was using gcc. Also find and
implement the CLOSE DATABASE workaround for bugs B42204 and
B64926 which lead to -1800 errors when a 7.x ESQL/C application
is talking to a 5.0x engine, also reported by Nathan.
97-05-29: Add support for $sth->rows and add simple test.
97-05-26: Add support for DBI->DataSources from DBI v0.82. Update required
software accordingly.
97-05-23: Fix Makefile.PL to recognize $INFORMIXDIR/bin at end of PATH,
thanks to Paul Kulchenko <paul@mega.donetsk.ua>.
97-05-20: Change the handling of deprecated attributes per Informix.pm.
Fix esqltest.ec when the same MODE ANSI database is used for both
connections (thanks to Kent S Gordon <kgor@inetspace.com> for
reporting this).
0.55 Bug-Fix Release
--------------------
97-05-20: ESQL/C Version 5.0x testing previously inadequate, and SQLSTATE
does not exist in 5.0x, sothe compilation fails. Also fixed the
core dumps from t/nulls01.t and t/nulls02.t (nasty workaround)
and the failure message in t/dbase.t. Tested with 5.06.UC1,
6.00.UE1, 7.22.UC2 and 9.10.UC2 on Sun Sparc 20 running Solaris
2.5.1, using 2 Perl configurations: Perl 5.003 with GCC 2.7.2.2
and Perl 5.004 with SUNWspro C Compiler SC3.0.1 -- all OK.
Version 0.55 released.
97-05-19: Change the handling of deprecated attributes per Informix.pm.
0.54 Release
------------
97-05-16: Another major rework on Informix.Licence to make the requirements
clearer to people with minimal exposure to Informix, after Honza
Pazdziora <adelton@fi.muni.cz> had some problems with it. Add
notes from Anibal Jodorcovsky <anibal@nortel.ca> about
compilation on Motorola M88K SVR4 machines in README.M88K.SVR4.
Version 0.54 released.
97-05-15: ChopBlanks code added and tested. VARCHAR was bad! Statement
level ChopBlanks attribute not tested (probably not operational;
attribute probably remains as it was when the statement was
prepared).
97-05-13: Finish adding driver fetch function, and consequential changes.
SQLSTATE now available via DBI::state!
97-05-05: Revise ixblob.ec to use ixblob.h and transfer to general library.
Eliminate BLOB_DUMMY_VALUE and BLOB_NULL_VALUE (can't think what
they were to be used for!). Consequential changes.
97-04-24: Add fix in Makefile.PL for DEC OSF/1 and ESQL/C 7.22 from Andrew
Hunt <andy@toolshed.com>. Reverse order of updates in ChangeLog.
DEC OSF/1 changes also reported later by Debbie Ingram
<djingram@mindspring.com>.
97-04-08: Add attribute $sth->{ix_Fetchable} to determine whether the
statement needs data fetching or not. Add this to sqlcmd.sh, which
enables basic simulation of the full SQLCMD program. Still need
better Getopt module which allows for multiple occurrences of a
single option on the command line (eg: multiple -e or -f options).
Also need better handling of filenames versus commands in the rest
of the arguments. Note that the build process does not (yet) do
anything with the source code of sqlcmd.sh.
97-04-01: Add '-Ae' for HP-UX unless $Config{ccflags} includes -Aa or -Ae,
per information from Oyvind Gjerstad <ogj@tglobe1.tollpost.no>.
Include $Config{ccflags} in the command line when building the
ESQL/C test program.
97-03-28: Correct the handling of $dbase2 in multiple connections tests when
DBD_INFORMIX_DATABASE2 is not set. Fix required after John C
Hossfeld <john@netpub.com> had problems. Also report which database
is being connected to! Problem (and solution) also subsequently
reported by James F Miner <jfm@winternet.com>.
97-03-26: Add ODBC type handling.
97-03-20: Update Makefile.PL to handle relocated DBIXS.h header, again.
97-03-18: Change the handling of deprecated attributes per Informix.PM.
0.53 Release
------------
97-03-06: Change the handling of deprecated attributes per Informix.PM.
Update the README file to spell out the pre-requisites better.
Update Makefile.PL with 0.53a1 version number, and so that it
insists that $INFORMIXDIR is set, and that $INFORMIXDIR/bin is in
$PATH. Update tests to use the non-deprecated attributes. Add the
test program esqltest.ec to check that the ESQL/C environment is OK
before creating the Makefile.
97-03-10: No changes over the weekend - the skiing at Squaw was fabulous!
Update the POD to encourage the use of DBI->connect() in preference
to $drh = DBI->install_driver(). Add two new, but fore-shadowed,
driver attributes, ix_CurrentConnection and ix_ActiveConnections,
though they're currently only set when the driver is loaded.
97-03-17: Update Informix.Licence to contain more comprehensive information
on licencing and Netscape's LiveWire Pro. Update t/updcursor.t so
it passes on a logged (non-MODE ANSI) database. Fix the problem
with nulls found by Oyvind Gjerstad <ogj@tglobe1.tollpost.no> and
now immortalized in the test case t/nulls02.t. Change version to
0.53 and release.
0.52 Release
------------
97-03-05: Add ESQL/C => DBD::Informix statement mapping to POD in Informix.PM,
after suggestion by Bill Mitlyng <Bill.Mitlyng@abnamro.com>.
97-03-03: Use $DBI::VERSION to determine where DBIXS.h should be found.
97-03-02: Split attribute handling code from dbdimp.ec into dbdattr.ec.
Add dbd_ix_module() function. Revise the interface to the functions
dbd_st_{FETCH|STORE} to match dbd_db_{FETCH|STORE} (they now take a
pointer to imp_sth_t argument instead of a pointer to SV).
97-02-28: Add the Informix.Licence file to explain the licencing issues for
Informix-ESQL/C. Update the README file. Fix the caching of
attributes. Add 'deprecated' warning to the system, and a way of
turning the deprecated warnings off. Use this method in
InformixTest.pm. Add the attribute handling of ix_XyZ as well as
deprecating the corresponding XyZ attribute. Fix the handling of
SQLCA so that the inserted serial number is available, etc. Version
bumped to 0.52.
0.51 Release
------------
97-02-26: Add $sth->{CursorName} and $dbh->{ConnectionName}. Add test
updcursor.t to check that you can do 'UPDATE ... WHERE CURRENT OF'.
Add ".DEFAULT." as an explicit way of making the default connection.
Release version 0.51, and a patch, version 0.51p1.
97-02-25: Version bumped to 0.51. Nulls on SELECT handled correctly. Test
nulls01.t added. Notes on access to the SQLCA record added to the
POD in Informix.PM, after this omission was noted by Torsten Stein
<stein@cls.de>.
97-01-23: Fix InformixTest.pm so it works with perldoc. It seems that perldoc
does not like lines which start with a single quote. Thanks to
Jason Bodnar <jcbodnar@mail.utexas.edu> for pointing this problem
out. Add and use stmt_err to report errors or warnings, and document
it.
0.50 Release
------------
97-02-13: Version 0.50 released. Handle null database names on connect so
that CREATE DATABASE, etc, can be used. ESQL/C 5.0x has to handle
this by doing nothing (successfully) in dbd_ix_database(), but the
disconnect code needs to know that there is no database so it
doesn't try to close the current database. This requires interface
changes for dbd_ix_database() and dbd_ix_connect(). Add testing.
Thanks to Serge Davidov <serge@fcmc.com> for finding the problem.
You can also use DBI->connect('@server','','','Informix'), which
actually worked in v0.25. DBD::Informix is still confused, in
general, by statements such as CLOSE DATABASE.
97-01-12: Version bumped to 0.50p1, for consistency with documented intentions
when 0.25 release was made.
0.26 Release
------------
97-01-12: Use cursors WITH HOLD when a MODE ANSI database is working with
AutoCommit set On. Make transaction tests more sensitive.
This release never made public.
97-01-09: Updated copyright messages. Improved documentation. Fixed tests.
Add $dbh->{InTransaction} query. Clean up dbdimp.ec so that
AutoCommit works better. Etc. Version 0.26a2.
96-12-20: Get transaction stuff to work, more or less, with AutoCommit On or
Off, in MODE ANSI or Logged databases. Tests t/transact0[123].t
added, but these still need to be more formally verified. They
pass with OnLine and any type of database. Clean up various tests
so that they pass on MODE ANSI databases. Add select_some_data
and select_zero_data functions to InformixTest.pm. Etc. Version
0.26a1. Use bug-fixed version of decsci.c (ouch!).
96-12-19: Add install instructions to README file after Wes Gamble
<wgamble@informix.com> pointed out this omission. Add extensive
notes to Informix.PM on transaction management. Implement most
of what those notes say (WITH HOLD cursors omitted at the
moment).
96-12-18: Fix bug in patched Makefile.PL (need space before __H_LOCALEDEF).
Updated dbd_ix_setconnection() to take connection information, and
updated code in dbdimp.ec to ensure that the correct connection is
set before starting work. Simple testing more or less works. Some
more major revisions, such as not passing SV* to most of the code in
dbdimp.ec (let the Informix.c code convert as much as possible to
plain C). Rework the 'do' function to completely override
DBI::_::db::do, avoiding problems with the rows() function.
0.25 Patch 2
------------
96-12-13: Update Makefile.PL to help the AIX people. Version 0.25p2 created
but only the Makefile.PL changes sent to the dbi-users@fugue.com so
not released officially, though available publicly.
0.25 Patch 1
------------
96-12-10: Handle 5.00 and 5.01 by supplying actual array names to GET
DESCRIPTOR, thus bypassing the bug. Note that it does affect 5.01
too; edited esql5_00.h to match. Patch sent to Jeff Rowe who
confirms it works. (Note to JL: 5.01 found on Godzilla; what is
installed as 5.00 is actually 4.10 ESQL/C). Version 0.25p1 created
but not released officially or publicly.
0.25 Release
------------
96-12-09: Finally worked out how SQL descriptors handle blobs. Code now works
safely with blobs located in memory. However, with ESQL/C 7.21.UC1
on Solaris 2.4, there is a memory leak corresponding to the space
allocated for the blobs for a given statement unless you manually
release it before deallocating the descriptor. This is nasty, and
may not be true of other versions. Added test t/blob02.t. Note
that t/blob03.t uses the data created by t/blob02.t, so it must be
run (reasonably successfully) before t/blob03.t can be run
successfully. Use decsci.[ch] code (plus auxilliary function
decgen() inside dbdimp.ec) to format FLOAT, SMALLFLOAT, DECIMAL and
MONEY quantities. The 5.0x and 6.0x libraries do not always handle
these gracefully.
With Thanks to the DBD::Informix v0.25 Testing Team:
John Wang - wangj@aromat.com
Del Simmons - del@infoman.com
Dave Thomas - dave@thomases.com
Oyvind Gjerstad - ogj@tglobe1.tollpost.no
Jeff Rowe - j.p.rowe@larc.nasa.gov
Rodlyn Joseph - bluephi@sharpcomm.com
Duc Nguyen - dhn@benvolio.nova.att.com
Serge Davidov - serge@fcmc.com
96-12-05: More revisions to Makefile.PL, some putting teeth into the
prerequisites, some more dealing with DEC OSF (there's a compiler
flag to say set __STDC__, as there usually is). Rename nm_* inside
the imp_sth structure. Redefine the type of imp_sth->blobs and
upgrade code to match. Test t/blob03.t passes; some work to do to
get t/blob04.t to pass. Improved README. Etc.
96-12-04: Update Makefile.PL to overcome Dave Thomas problems on DEC OSF
(version 25.13 sent to him for approval, and comments incorporated
into version 25.14 .. 25.16). Also finish converting the tests in
the t directory.
96-12-03: Update Makefile.PL (and esqlcc) to use '$(FULLPERL) esqlcc' etc.
Rename the extratests directory to t for automatic testing as
documented in the Test::Harness module. Create (and document)
InformixTest.pm to help with the test programs. Note that
test.pl does not exploit it all that much because it has to test
some more basic stuff. However, the other tests do use it.
NB: ESQL/C 5.00 and 5.01 have a bug in the SQL descriptor code
which means that using 'char *' as a type restricts you to precisely
3 characters of data. This is fixed in 5.02 and above. Dammit!
96-12-02: Release 0.25b3 with some work incomplete. Update Makefile.PL so
that "make perl" works. Note that INSERT and blobs does not work!
96-12-01: Upgrade Makefile.PL to use new Perl scripts esqlcc, esqlsed and
esqlld to protect platforms such as DEC OSF/1 where the C
compiler is not used to link shared libraries. Also, only use
GCC warnings when DBD_GCC_DEBUG is set in the environment. Also
use $(RM_F) to remove the generated C files instead of $(RM) so
that it works for platforms like SCO where $(RM) is not defined
by MAKE automatically.
96-11-30: Move sqlca handling to connection (not statement).
Fix blob handling to the extent that SELECT works. (INSERT and
UPDATE still need fixing. INSERT will probably be able to adapt
the SELECT code to handle blobs. UPDATE will be very tricky.
Memory leaks may (probably) occur.
Upgrade Makefile.PL to always set INFORMIXC when compiling ESQL/C.
96-11-26: Feedback from beta testers incorporated into build process.
Not all problems will be resolved yet. Version 0.25b2.
96-11-25: Changes leading up to 0.25b1 release. Add sqlca record handling
for statements. Add use of EXEC IMMEDIATE when feasible.
96-11-19: Rewrite of dbdimp.ec more or less complete. The ESQL/C code
has been almost totally rewritten; the Perl interface code is
still similar to what was there previously. Nearly ready to go
into release. 5.0x ESQL/C not linking OK on Solaris; not sure
why. 7.21 is doing fine.
96-11-01: Control transferred to Jonathan Leffler <johnl@informix.com>.
0.24 Release
------------
96-09-26: Patches from Terry arrived. This'll be the 0.25 release. Seems to
work.
0.23 Release
------------
96-09-18: Wrapped in a patch to Makefile.PL from Christophe Martin. Added
some other Makefile.PL patches of my own. It should now support
cleaner builds on HP-UX (spit!) and correctly detect when dbdimp.ec
has changed.
96-09-18: Wrapped in patches by Terry Nightingale based on Bill Hailes'
isqlperl for supporting multi-cursors. Cleaned the type handling
code again.
96-09-17: Added patch for bad column names as show in George Vicherek's
splendid test case. (extratests/primitives.pl exercises this)
0.22 Release
------------
96-09-10: Fixed all numeric types to return correctly. CHAR fields don't
quite behave as planned, or perhaps they do. Added 'multicursor'
test which proves that simultaneous multiple cursors fail miserably.
96-09-09: Added @ary = $drh->func('_ListDBs'); method for getting a list of
extant databases on the local Informix server. (From original code
and suggestion by George.Vicherek@Sciatl.COM)
0.21 Release
------------
95-09-01: Fixed truncated INTEGER bug. Other numerical types still screwed.
Patch on the way.
Pre-Release
-----------
96-04-14: Added support for CREATE, DROP, INSERT, UPDATE and DELETE. Fixed
$DBI::errnum and $DBI::errstr problems. Fixed Informix.pm to stop
reporting MSQL_HOME errors. Fixed Makefile.PL. Removed hacky
'build' file and uses Makefile.PL exclusively now.
95-10-17: After a good break of about a month and a half, back in the saddle.
We have data from SELECTs coming back from the database! Released
0.20pl0 Disney on ftp.mcqueen.com. Limited hacker release.
95-08-18: Started. Version 0.0.0 (Totally pre-alpha!)
--------------
Jonathan Leffler <johnl@informix.com>
@(#)$Id: ChangeLog,v 59.5 1998/03/11 17:43:01 jleffler Exp $