]> granicus.if.org Git - postgresql/log
postgresql
24 years agoMore updates for function call interface > 8.
Bruce Momjian [Mon, 10 Jan 2000 18:27:41 +0000 (18:27 +0000)]
More updates for function call interface > 8.

24 years agoUpdate fmgr to allow 32 arguments.
Bruce Momjian [Mon, 10 Jan 2000 18:18:30 +0000 (18:18 +0000)]
Update fmgr to allow 32 arguments.

24 years agoMake number of args to a function configurable.
Bruce Momjian [Mon, 10 Jan 2000 17:14:46 +0000 (17:14 +0000)]
Make number of args to a function configurable.

24 years agoRename oid8 -> oidvector and int28 -> int2vector. Cleanup of *out functions.
Bruce Momjian [Mon, 10 Jan 2000 16:13:23 +0000 (16:13 +0000)]
Rename oid8 -> oidvector and int28 -> int2vector.  Cleanup of *out functions.

24 years agoUpdate int28out and out8out and _in_ functions to handle trailing zeros
Bruce Momjian [Mon, 10 Jan 2000 15:41:34 +0000 (15:41 +0000)]
Update int28out and out8out and _in_ functions to handle trailing zeros
properly.

24 years agoUpdate TODO list.
Bruce Momjian [Mon, 10 Jan 2000 13:20:57 +0000 (13:20 +0000)]
Update TODO list.

24 years agoImprove cache invalidation handling. Eespecially
Hiroshi Inoue [Mon, 10 Jan 2000 06:30:56 +0000 (06:30 +0000)]
Improve cache invalidation handling. Eespecially
this would fix TODO
* elog() flushes cache, try invalidating just entries from
  current xact, perhaps using invalidation cache

24 years agoFix oid8in and int28in for spaces
Bruce Momjian [Mon, 10 Jan 2000 05:23:47 +0000 (05:23 +0000)]
Fix oid8in and int28in for spaces

24 years agoMove fixes for >8 indexed fields.
Bruce Momjian [Mon, 10 Jan 2000 05:20:26 +0000 (05:20 +0000)]
Move fixes for >8 indexed fields.

24 years agoMove INDEX_MAX_KEYS to postgres.h, and make it configurable for users.
Bruce Momjian [Mon, 10 Jan 2000 04:36:37 +0000 (04:36 +0000)]
Move INDEX_MAX_KEYS to postgres.h, and make it configurable for users.

24 years agoUpdate TODO list.
Bruce Momjian [Mon, 10 Jan 2000 04:23:58 +0000 (04:23 +0000)]
Update TODO list.

24 years agoRepair subtle VACUUM bug that led to 'HEAP_MOVED_IN was not expected'
Tom Lane [Mon, 10 Jan 2000 04:09:50 +0000 (04:09 +0000)]
Repair subtle VACUUM bug that led to 'HEAP_MOVED_IN was not expected'
errors.  VACUUM normally compacts the table back-to-front, and stops
as soon as it gets to a page that it has moved some tuples onto.
(This logic doesn't make for a complete packing of the table, but it
should be pretty close.)  But the way it was checking whether it had
got to a page with some moved-in tuples was to look at whether the
current page was the same as the last page of the list of pages that
have enough free space to be move-in targets.  And there was other
code that would remove pages from that list once they got full.
There was a kluge that prevented the last list entry from being
removed, but it didn't get the job done.  Fixed by keeping a separate
variable that contains the largest block number into which a tuple
has been moved.  There's no longer any need to protect the last element
of the fraged_pages list.
Also, fix NOTICE messages to describe elapsed user/system CPU time
correctly.

24 years agoUpdate platform-specific-expected-file support so that platforms can be
Tom Lane [Sun, 9 Jan 2000 20:54:36 +0000 (20:54 +0000)]
Update platform-specific-expected-file support so that platforms can be
specified by regular-expression patterns.  Add some more files.

24 years agoinstall_plpgsql is no longer a regress test (it's done via createlang);
Tom Lane [Sun, 9 Jan 2000 19:34:43 +0000 (19:34 +0000)]
install_plpgsql is no longer a regress test (it's done via createlang);
remove the no-longer-used files.

24 years agoThe psql online help for ALTER TABLE (\h alter table) is corrupt. I
Bruce Momjian [Sun, 9 Jan 2000 17:35:27 +0000 (17:35 +0000)]
The psql online help for ALTER TABLE (\h alter table) is corrupt.  I
traced this back to what I believe is an error in the sgml file used to
generate this comment, found in pgsql/doc/src/sgml/ref/alter_table.sgml.

Stephen Birch

24 years agoAdd SetPidFile() and friends.
Tatsuo Ishii [Sun, 9 Jan 2000 12:19:27 +0000 (12:19 +0000)]
Add SetPidFile() and friends.

24 years agoDo not start if postmaster is running.
Tatsuo Ishii [Sun, 9 Jan 2000 12:17:33 +0000 (12:17 +0000)]
Do not start if postmaster is running.

24 years agoMove SetPidFile() and firends to utils/init/miscinit.c from
Tatsuo Ishii [Sun, 9 Jan 2000 12:15:57 +0000 (12:15 +0000)]
Move SetPidFile() and firends to utils/init/miscinit.c from
postmaster/postmaster.c so that
tcop/postgres.c can use them. Now we have an interlock between
postmaster and postgres.

24 years agoMove SetPidFile() and firends to utils/init/miscinit.c so that
Tatsuo Ishii [Sun, 9 Jan 2000 12:13:24 +0000 (12:13 +0000)]
Move SetPidFile() and firends to utils/init/miscinit.c so that
tcop/postgres.c can use them. Now we have an interlock between
postmaster and postgres.

24 years agoAdd more portability to echo -n (code stolen from createlang)
Tatsuo Ishii [Sun, 9 Jan 2000 12:06:52 +0000 (12:06 +0000)]
Add more portability to echo -n (code stolen from createlang)
Do not start postmaster if postgres is running

24 years agoNew scheme for managing platform-specific regress test result files.
Tom Lane [Sun, 9 Jan 2000 07:54:00 +0000 (07:54 +0000)]
New scheme for managing platform-specific regress test result files.
Instead of hard-wiring one result file per platform, there is a map file
'resultmap' that says which one to use --- a lot like template/.similar.
I have only created entries in resultmap for my own platform (HPUX) so
far; feel free to add lines for other platforms.

24 years agoRemove obsolete platform-specific comparison files.
Tom Lane [Sun, 9 Jan 2000 06:34:46 +0000 (06:34 +0000)]
Remove obsolete platform-specific comparison files.

24 years agoFirst examples of multiplatform result comparison files.
Tom Lane [Sun, 9 Jan 2000 06:32:33 +0000 (06:32 +0000)]
First examples of multiplatform result comparison files.

24 years agoRemove obsolete platform-specific regress test comparison files.
Tom Lane [Sun, 9 Jan 2000 06:30:55 +0000 (06:30 +0000)]
Remove obsolete platform-specific regress test comparison files.
Note: don't put any of these back till you've grokked the new code for
platform-specific comparisons that I'm about to commit...

24 years agoRemove CVS $Header lines from a couple of regress test files that had
Tom Lane [Sun, 9 Jan 2000 04:01:49 +0000 (04:01 +0000)]
Remove CVS $Header lines from a couple of regress test files that had
them --- it is just *way* too painful to keep expected results in sync
when these are present.

24 years agoUpdate remaining tests for new psql, with the exception of 'arrays'.
Tom Lane [Sun, 9 Jan 2000 03:50:27 +0000 (03:50 +0000)]
Update remaining tests for new psql, with the exception of 'arrays'.

24 years agoUpdate remaining tests for new psql, with the exception of 'arrays',
Tom Lane [Sun, 9 Jan 2000 03:48:39 +0000 (03:48 +0000)]
Update remaining tests for new psql, with the exception of 'arrays',
which is broken in some weird way that I don't understand.  I think it
may be exposing a bug in the new psql --- for one thing, I get different
results when I run psql by hand than the regress script gets.  What
the heck???

24 years agoFix some missing substitutions of _OBJWD_ and _DLSUFFIX_.
Tom Lane [Sun, 9 Jan 2000 02:48:01 +0000 (02:48 +0000)]
Fix some missing substitutions of _OBJWD_ and _DLSUFFIX_.

24 years agoAnother round of planner/optimizer work. This is just restructuring and
Tom Lane [Sun, 9 Jan 2000 00:26:47 +0000 (00:26 +0000)]
Another round of planner/optimizer work.  This is just restructuring and
code cleanup; no major improvements yet.  However, EXPLAIN does produce
more intuitive outputs for nested loops with indexscans now...

24 years agoThis patch removes the initialization of ri in loop in
Bruce Momjian [Sat, 8 Jan 2000 22:58:03 +0000 (22:58 +0000)]
This patch removes the initialization of ri in loop in
quote_postgres(...) in ecpglib.c.

The code in CVS reads:

quote_postgres(char *arg, int lineno)
{
 char    *res = (char *) ecpg_alloc(2 * strlen(arg) + 3, lineno);
 int   i,
    ri = 0;

 if (!res)
  return (res);

 res[ri++] = '\'';
 for (i = 0, ri=0; arg[i]; i++, ri++)
 {
  switch (arg[i])
  {
   case '\'':
    res[ri++] = '\'';
    break;
   case '\\':
    res[ri++] = '\\';
    break;
   default:
    ;
  }

The problem here is that ri is reset to 0, thus overwriting the initial
quote.

Stephen Birch

24 years agoModify PageIsEmpty and PageGetMaxOffsetNumber macros to behave sanely
Tom Lane [Sat, 8 Jan 2000 21:59:55 +0000 (21:59 +0000)]
Modify PageIsEmpty and PageGetMaxOffsetNumber macros to behave sanely
if presented an uninitialized (all zeroes) page.  The system no longer
crashes hard if an all-zeroes page is present in a relation.  There seem
to be some boundary conditions where a page will be appended to a relation
and zeroed, but its page header is never initialized; until we can track
down and fix all of those, robustness seems like a good idea.
Also, clean up some obsolete and downright wrong comments.

24 years agoNeed defense against oversize index entries in btree CREATE INDEX,
Tom Lane [Sat, 8 Jan 2000 21:24:49 +0000 (21:24 +0000)]
Need defense against oversize index entries in btree CREATE INDEX,
as well as when inserting entries into an existing index.

24 years agoSorry, that I send this letter/patch again, but previous sending is
Bruce Momjian [Fri, 7 Jan 2000 17:22:47 +0000 (17:22 +0000)]
Sorry, that I send this letter/patch again, but previous sending is
still
without answer. I want continue with to_char(), but I need any answer
for this patch. Please.

Thank! (and sorry of my impatient :-)
                                                        Karel

24 years agoCorrect grammatical error
Tatsuo Ishii [Fri, 7 Jan 2000 09:28:03 +0000 (09:28 +0000)]
Correct grammatical error

24 years ago*** empty log message ***
Michael Meskes [Fri, 7 Jan 2000 08:01:18 +0000 (08:01 +0000)]
*** empty log message ***

24 years agoUpdate pg_dumpall for new psql format.
Bruce Momjian [Thu, 6 Jan 2000 21:16:18 +0000 (21:16 +0000)]
Update pg_dumpall for new psql format.

24 years agoChanged "triggered data change violation" detection code
Jan Wieck [Thu, 6 Jan 2000 20:47:01 +0000 (20:47 +0000)]
Changed "triggered data change violation" detection code
in trigger manager.

Jan

24 years agoFixed bug in targetlist expression replacement of
Jan Wieck [Thu, 6 Jan 2000 16:30:43 +0000 (16:30 +0000)]
Fixed bug in targetlist expression replacement of
SET DEFAULT referential action triggers.

Jan

24 years agoClean up header for uniform appearance throughout tests.
Thomas G. Lockhart [Thu, 6 Jan 2000 06:41:55 +0000 (06:41 +0000)]
Clean up header for uniform appearance throughout tests.

24 years agoUpdate for new psql formatting.
Thomas G. Lockhart [Thu, 6 Jan 2000 06:40:54 +0000 (06:40 +0000)]
Update for new psql formatting.

24 years agoFreshen up the banner displayed when running the regression test.
Thomas G. Lockhart [Thu, 6 Jan 2000 06:40:18 +0000 (06:40 +0000)]
Freshen up the banner displayed when running the regression test.

24 years agoFix it's and its to be correct.
Bruce Momjian [Wed, 5 Jan 2000 18:23:54 +0000 (18:23 +0000)]
Fix it's and its to be correct.

24 years agoUpdate format to add uniform headers on files.
Thomas G. Lockhart [Wed, 5 Jan 2000 17:32:29 +0000 (17:32 +0000)]
Update format to add uniform headers on files.

24 years agoUpdate format to add uniform headers on files.
Thomas G. Lockhart [Wed, 5 Jan 2000 17:32:18 +0000 (17:32 +0000)]
Update format to add uniform headers on files.
Update output to new psql conventions.

24 years agoUpdate output to new psql conventions.
Thomas G. Lockhart [Wed, 5 Jan 2000 17:31:08 +0000 (17:31 +0000)]
Update output to new psql conventions.

24 years agoFix spaces in text message.
Thomas G. Lockhart [Wed, 5 Jan 2000 17:30:05 +0000 (17:30 +0000)]
Fix spaces in text message.

24 years agoClean up format of tests.
Thomas G. Lockhart [Wed, 5 Jan 2000 06:07:58 +0000 (06:07 +0000)]
Clean up format of tests.
Remove older "::" type coersion syntax in favor of extended SQL92 style.
Include a few new tests for datetime/timespan arithmetic.

24 years agoVerified output from new psql.
Thomas G. Lockhart [Wed, 5 Jan 2000 06:06:23 +0000 (06:06 +0000)]
Verified output from new psql.
Include a few new tests for datetime/timespan arithmetic.

24 years agoMove numeric test to be near other numeric data types like int4 and float8.
Thomas G. Lockhart [Wed, 5 Jan 2000 06:04:56 +0000 (06:04 +0000)]
Move numeric test to be near other numeric data types like int4 and float8.

24 years agoClean up syntax to use SQL92-ish type coersion
Thomas G. Lockhart [Tue, 4 Jan 2000 16:21:02 +0000 (16:21 +0000)]
Clean up syntax to use SQL92-ish type coersion
 rather than the Postgres "::" notation.
All of these tests have been completely inspected and give correct results.

24 years agoMatch results with format from new psql.
Thomas G. Lockhart [Tue, 4 Jan 2000 16:19:34 +0000 (16:19 +0000)]
Match results with format from new psql.
All of these tests have been completely inspected and give correct results.

24 years agoRepair two recently reported problems:
Thomas G. Lockhart [Tue, 4 Jan 2000 07:53:27 +0000 (07:53 +0000)]
Repair two recently reported problems:
1) datetime_pl_span() added the seconds field before adding the months
 field.  This lead to erroneous results for e.g.
   select datetime '1999-11-30' + timespan '1 mon - 1 sec';
 Reverse the order of operations to add months first.
2) tm2timespan() did all intermediate math as integer, converting to double
 at the very end. This resulted in hidden overflows when given very large
 integer days, hours, etc. For example,
   select '74565 days'::timespan;
 produced the wrong result. Change code to ensure that doubles are used
 for intermediate calculations.
Thanks to Olivier PRENANT <ohp@pyrenet.fr> and
 Tulassay Zsolt <zsolt@tek.bke.hu> for problem reports and to Tom Lane for
 accurate analyses.

24 years agoUpdate DATEDEBUG removal.
Bruce Momjian [Sun, 2 Jan 2000 02:32:37 +0000 (02:32 +0000)]
Update DATEDEBUG removal.

24 years agoUpdate length of timestamp to 30.
Bruce Momjian [Sun, 2 Jan 2000 02:11:08 +0000 (02:11 +0000)]
Update length of timestamp to 30.

24 years agoRemove DATEDEBUG because it didn't look Y2K safe, and fix timestamp elog
Bruce Momjian [Sun, 2 Jan 2000 01:37:28 +0000 (01:37 +0000)]
Remove DATEDEBUG because it didn't look Y2K safe, and fix timestamp elog
to be Y2K safe.

24 years agoMake psql \dT always show descriptions. There is room.
Bruce Momjian [Fri, 31 Dec 1999 21:06:19 +0000 (21:06 +0000)]
Make psql \dT always show descriptions.  There is room.

24 years agoGenerate double-sided LIKE indexquals that work even in weird locales,
Tom Lane [Fri, 31 Dec 1999 05:38:25 +0000 (05:38 +0000)]
Generate double-sided LIKE indexquals that work even in weird locales,
by continuing to increment the rightmost character until we get a string
that is demonstrably greater than the pattern prefix.

24 years agoClean up loose end in LIKE optimization fix: parser's code would generate
Tom Lane [Fri, 31 Dec 1999 03:41:03 +0000 (03:41 +0000)]
Clean up loose end in LIKE optimization fix: parser's code would generate
<= and >= indexquals from a LIKE even if the index in question didn't
support those operators.  (As, for example, a hash index does not.)

24 years agoClean up datatypes and comments for op_class() routine.
Tom Lane [Fri, 31 Dec 1999 03:18:43 +0000 (03:18 +0000)]
Clean up datatypes and comments for op_class() routine.

24 years agoRevise init_sequence so that it doesn't leak memory if the requested
Tom Lane [Fri, 31 Dec 1999 00:54:27 +0000 (00:54 +0000)]
Revise init_sequence so that it doesn't leak memory if the requested
sequence doesn't exist.

24 years agoelog() was set up to call abort() if it saw an ERROR or FATAL exit
Tom Lane [Thu, 30 Dec 1999 23:03:40 +0000 (23:03 +0000)]
elog() was set up to call abort() if it saw an ERROR or FATAL exit
during InitProcessingMode and the CurrentTransactionState was neither
TRANS_DEFAULT nor TRANS_DISABLED.  Unfortunately, after someone's recent
change to start the transaction manager earlier in startup than it used
to be started, that caused an abort() and consequent database system
reset on quite harmless errors (such as rejecting an invalid user name!).
As far as I can see, the test on CurrentTransactionState was completely
useless anyway, so I've removed it.

24 years agoClearify DECLARE syntax by saying cursorname, not just cursor.
Bruce Momjian [Thu, 30 Dec 1999 22:58:10 +0000 (22:58 +0000)]
Clearify DECLARE syntax by saying cursorname, not just cursor.

24 years agoAllow --with-mb=SQL_ASCII
Tatsuo Ishii [Thu, 30 Dec 1999 08:33:49 +0000 (08:33 +0000)]
Allow --with-mb=SQL_ASCII

24 years agoRepair bugs discussed in pghackers thread of 15 May 1999: creation of a
Tom Lane [Thu, 30 Dec 1999 05:05:13 +0000 (05:05 +0000)]
Repair bugs discussed in pghackers thread of 15 May 1999: creation of a
relcache entry no longer leaks a small amount of memory.  index_endscan
now releases all the memory acquired by index_beginscan, so callers of it
should NOT pfree the scan descriptor anymore.

24 years agoImprove subquery error message, now says "More than one tuple returned
Bruce Momjian [Wed, 29 Dec 1999 22:57:17 +0000 (22:57 +0000)]
Improve subquery error message, now says "More than one tuple returned
by subselect used as expression."

24 years agoMove ipc patch into README.NT.
Bruce Momjian [Wed, 29 Dec 1999 10:28:00 +0000 (10:28 +0000)]
Move ipc patch into README.NT.

24 years agoAdd NT patch.
Bruce Momjian [Wed, 29 Dec 1999 10:24:16 +0000 (10:24 +0000)]
Add NT patch.

24 years agoMove NT patch into readme.
Bruce Momjian [Wed, 29 Dec 1999 10:18:04 +0000 (10:18 +0000)]
Move NT patch into readme.

24 years agoUpdate comment.
Bruce Momjian [Wed, 29 Dec 1999 10:13:20 +0000 (10:13 +0000)]
Update comment.

24 years agoUpdate comments.
Bruce Momjian [Wed, 29 Dec 1999 10:12:23 +0000 (10:12 +0000)]
Update comments.

24 years agoRemoved LZTEXT datatype as discussed.
Jan Wieck [Tue, 28 Dec 1999 13:40:53 +0000 (13:40 +0000)]
Removed LZTEXT datatype as discussed.

Jan

24 years agoFix length limit, MikeA
Bruce Momjian [Mon, 27 Dec 1999 18:21:07 +0000 (18:21 +0000)]
Fix length limit, MikeA

24 years agoHi, all,
Bruce Momjian [Mon, 27 Dec 1999 15:45:04 +0000 (15:45 +0000)]
Hi, all,

This is the patch for the final bit.  Sorry that it's separate.

Cheers...

MikeA

24 years agoHi, all
Bruce Momjian [Mon, 27 Dec 1999 15:42:44 +0000 (15:42 +0000)]
Hi, all

I finally got around to schlepping through pg_dump, to finish what I started
about three months (or more) ago.  Attached is a gzipped diff file to apply
in the bin/pg_dump directory.  This should remove all string length
dependencies, except one, which I'm working on.  It has been through some
rudimentary unit testing, but that's about it, so if any of you would give
it a more strenuous run-through, I'd be grateful for the feedback.

Cheers...

 Ansley, Michael

24 years agoPut back erroneously-removed definition of 'defines' variable.
Tom Lane [Sun, 26 Dec 1999 21:31:35 +0000 (21:31 +0000)]
Put back erroneously-removed definition of 'defines' variable.

24 years agoIt turns out that the item size limit for btree indexes is about BLCKSZ/3,
Tom Lane [Sun, 26 Dec 1999 03:48:22 +0000 (03:48 +0000)]
It turns out that the item size limit for btree indexes is about BLCKSZ/3,
not BLCKSZ/2 as some of us thought.  Add check for oversize item so that
failure is detected before corrupting the index, not after.

24 years agoUpdate developers faq.
Bruce Momjian [Fri, 24 Dec 1999 16:46:11 +0000 (16:46 +0000)]
Update developers faq.

24 years agoClean up handling of explicit NULL constants. Cases like
Tom Lane [Fri, 24 Dec 1999 06:43:34 +0000 (06:43 +0000)]
Clean up handling of explicit NULL constants.  Cases like
SELECT null::text;
SELECT int4fac(null);
work as expected now.  In some cases a NULL must be surrounded by
parentheses:
SELECT 2 + null;                 fails
SELECT 2 + (null);               OK
This is a grammatical ambiguity that seems difficult to avoid.  Other
than that, NULLs seem to behave about like you'd expect.  The internal
implementation is that NULL constants are typed as UNKNOWN (like
untyped string constants) until the parser can deduce the right type.

24 years ago*** empty log message ***
Michael Meskes [Thu, 23 Dec 1999 12:33:19 +0000 (12:33 +0000)]
*** empty log message ***

24 years agoFix minor bug.
Tatsuo Ishii [Wed, 22 Dec 1999 04:41:17 +0000 (04:41 +0000)]
Fix minor bug.

24 years agoAdd installtion of postmaster.opts.default.
Tatsuo Ishii [Wed, 22 Dec 1999 04:23:31 +0000 (04:23 +0000)]
Add installtion of postmaster.opts.default.

24 years agoAdd installation of pg_ctl
Tatsuo Ishii [Wed, 22 Dec 1999 04:12:55 +0000 (04:12 +0000)]
Add installation of pg_ctl
Locate path of postmaster in a portable way (stolen from initdb)
Add postmaster.opts.default.sample which should be copied into
$PGLIB in the installtion process. Also, it will be installed into
$PGDATA while initdb is running.

24 years agoto live in a transaction before access to db
Hiroshi Inoue [Wed, 22 Dec 1999 00:07:16 +0000 (00:07 +0000)]
to live in a transaction before access to db
during backend startup.

24 years agoupdate_pg_pwd() is an AR trigger. Corrected return type.
Jan Wieck [Tue, 21 Dec 1999 22:39:02 +0000 (22:39 +0000)]
update_pg_pwd() is an AR trigger. Corrected return type.

Jan

24 years agoThe first fix is to allow an input file with a relative path and without
Bruce Momjian [Tue, 21 Dec 1999 17:42:16 +0000 (17:42 +0000)]
The first fix is to allow an input file with a relative path and without
a ".pgc " extension. The second patch fixes a coredump when there is
more than one input file (in that case, cur and types were not set to
NULL before processing the second f ile)

The patch below modifies the accepted grammar of ecpg to accept

 FETCH [direction] [amount] cursor name

i.e. the IN|FROM clause becomes optional (as in Oracle and Informix).
This removes the incompatibility mentioned in section "Porting From
Other RDBMS Packages" p169, PostgreSQL Programmer's Guide. The grammar
is modified in such a way as to avoid shift/reduce conflicts. It does
not accept the statement "EXEC SQL FETCH;" anymore, as the old grammar
did (this seems to be a bug of the old grammar anyway).

This patch cleans up the handling of space characters in the scanner;
some patte rns require \n to be in {space}, some do not. A second fix is
the handling of cpp continuati on lines; the old pattern did not match
these. The parser is patched to fix an off-by-one error in the #line
directives. The pa rser is also enhanced to report the correct location
of errors in declarations in the "E XEC SQL DECLARE SECTION". Finally,
some right recursions in the parser were replaced by  left-recursions.

This patch adds preprocessor directives to ecpg; in particular

EXEC SQL IFDEF, EXEC SQL IFNDEF, EXEC SQL ELSE, EXEC SQL ELIF and EXEC SQL ENDIF

"EXEC SQL IFDEF" is used with defines made with "EXEC SQL DEFINE" and
defines, specified on the command line with -D. Defines, specified on
the command line are persistent across multiple input files. Defines can
be nested up to a maximum level of 128 (see patch). There is a fair
amount of error checking to make sure directives are matched properly. I
need preprocessor directives for porting code, that is written for an
Informix database, to a PostgreSQL database, while maintaining
compatibility with the original code. I decided not to extend the
already large ecpg grammar. Everything is done in the scanner by adding
some states, e.g. to skip all input except newlines and directives. The
preprocessor commands are compatible with Informix. Oracle uses a cpp
replacement.

Rene Hogendoorn

24 years agoUpdate developers faq in main tree.
Bruce Momjian [Tue, 21 Dec 1999 17:34:05 +0000 (17:34 +0000)]
Update developers faq in main tree.

24 years agoThis patch will avoid SIGFPE on some geo functions , if PostgreSQL is compiled
Bruce Momjian [Tue, 21 Dec 1999 17:01:44 +0000 (17:01 +0000)]
This patch will avoid SIGFPE on some geo functions , if PostgreSQL is compiled
with DEC C.

DEC C doesn't handle double values greater than DBL_MAX, but some
PostgreSQL geo functions assign greater than DBL_MAX values to some vars
in some special cases - that couses SIGFPE. I dunno if that is the only place
to fix to work well with DEC C.

Kirill Nosov.

24 years agoautoconf
Bruce Momjian [Tue, 21 Dec 1999 16:56:42 +0000 (16:56 +0000)]
autoconf

24 years agoClean up qnx template finding.
Bruce Momjian [Tue, 21 Dec 1999 16:55:38 +0000 (16:55 +0000)]
Clean up qnx template finding.

24 years agoAdded empty TOASTER files and corrected some minor glitches
Jan Wieck [Tue, 21 Dec 1999 00:06:44 +0000 (00:06 +0000)]
Added empty TOASTER files and corrected some minor glitches
in regression tests.

Jan

24 years agotr cleanup
Bruce Momjian [Mon, 20 Dec 1999 14:41:55 +0000 (14:41 +0000)]
tr cleanup

24 years agoRequired catalog changes for extended LONG attribute storage.
Jan Wieck [Mon, 20 Dec 1999 10:40:43 +0000 (10:40 +0000)]
Required catalog changes for extended LONG attribute storage.

Jan

24 years agoCleanup of effective username test.
Bruce Momjian [Mon, 20 Dec 1999 05:39:40 +0000 (05:39 +0000)]
Cleanup of effective username test.

24 years agoFinally found a platform which has finite() but nonetheless sets errno
Tom Lane [Mon, 20 Dec 1999 02:15:35 +0000 (02:15 +0000)]
Finally found a platform which has finite() but nonetheless sets errno
rather than returning a NaN for bogus input to pow().  Namely, HPUX 10.20.
I think this is sufficient evidence for what I thought all along, which
is that the float.c code *must* look at errno whether finite() exists or
not.

24 years agoClean up some minor gcc warnings.
Tom Lane [Mon, 20 Dec 1999 01:41:32 +0000 (01:41 +0000)]
Clean up some minor gcc warnings.

24 years agoClean up some minor gcc warnings.
Tom Lane [Mon, 20 Dec 1999 01:34:19 +0000 (01:34 +0000)]
Clean up some minor gcc warnings.

24 years agoClean up some minor gcc warnings. I'm not touching the
Tom Lane [Mon, 20 Dec 1999 01:31:26 +0000 (01:31 +0000)]
Clean up some minor gcc warnings.  I'm not touching the
major one, though, which is the truly ugly stores into libpq private
storage.  Can't you find a better way to do this?

24 years agoClean up some minor gcc warnings.
Tom Lane [Mon, 20 Dec 1999 01:23:04 +0000 (01:23 +0000)]
Clean up some minor gcc warnings.

24 years agoWhoever touched this code last doesn't seem to understand
Tom Lane [Mon, 20 Dec 1999 01:19:58 +0000 (01:19 +0000)]
Whoever touched this code last doesn't seem to understand
what a header file is for :-(

24 years agoAvoid compiler warnings on systems that have snprintf and/or vsnprintf
Tom Lane [Mon, 20 Dec 1999 00:51:25 +0000 (00:51 +0000)]
Avoid compiler warnings on systems that have snprintf and/or vsnprintf
but do not bother to declare them in <stdio.h>.  Seems to be a more
common omission than you'd think...