I got a note from someone about problems installing Postgres 6.4
on HPUX, one of which was that HP's yacc wouldn't process gram.y
due to table overflow. That's hardly news ... but a distribution
tarball shouldn't require gram.y to be processed; it should contain
usable output files, no?
Looking into it, I find that the 6.4 tarball contains files with the
1998-10-29 23:54 postgresql-v6.4/src/backend/parser/gram.c
1998-10-14 11:56 postgresql-v6.4/src/backend/parser/gram.y
1998-09-30 01:48 postgresql-v6.4/src/backend/parser/parse.h
Since parser/Makefile has the dependency
gram.c parse.h: gram.y
$(YACC) $(YFLAGS) $<
the fact that parse.h is back-dated means that installers of 6.4 will
have to process gram.y.
In short, parse.h needs to be 'touch'ed in the repository.
I did that for the REL6_4 branch, but I wanted to raise a flag here
for updaters of the grammar: make sure that parse.h gets committed
when gram.c does. You may need to use "cvs commit -f" to force a
commit even though parse.h hasn't changed ... that looks to be the
cause of this particular glitch.
regards, tom lane
pgsql-hackers by date
|Next:||From: The Hermit Hacker||Date: 1998-11-22 22:44:29|
|Subject: Re: [HACKERS] What happened to FAQ_Linux, FAQ_Irix?|
|Previous:||From: Tom Lane||Date: 1998-11-22 21:50:26|
|Subject: What happened to FAQ_Linux, FAQ_Irix?|