Thread: Re: [COMMITTERS] pgsql: Don't need -Wno-error anymore, because flex is no longer

Peter Eisentraut wrote:
> Log Message:
> -----------
> Don't need -Wno-error anymore, because flex is no longer producing warnings.

I see this patch only affects ecpg?  Is this related to Tom's report
from trying the new flex?  Are we assuming that all CVS build people
also have the new flex?  Is the new flex even installed on the server
that makes our tarballs?  Are we all supposed to use the new flex?

I am confused.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://postgres.enterprisedb.com
 + If your life is a hard drive, Christ can be your backup. +

Am Dienstag, 18. März 2008 schrieb Bruce Momjian:
> Peter Eisentraut wrote:
> > Log Message:
> > -----------
> > Don't need -Wno-error anymore, because flex is no longer producing
> > warnings.
>
> I see this patch only affects ecpg?  Is this related to Tom's report
> from trying the new flex?

Yes.  There used to be a few -Wno-error uses throughout the source code, near 
the flex calls, to work around the well-known flex warnings issue in case 
someone wanted to use -Werror to check his builds.  The fact that there was 
only one such use left in the ecpg subtree shows that this wasn't really 
maintained and could never have worked reliably.

> Are we assuming that all CVS build people 
> also have the new flex?  Is the new flex even installed on the server
> that makes our tarballs?  Are we all supposed to use the new flex?

I think it is reasonable to assume that those wanting to use -Werror should 
use the new flex, because if they don't the build will fail anyway, at any of 
the other lexers.

Peter Eisentraut <peter_e@gmx.net> writes:
> I think it is reasonable to assume that those wanting to use -Werror should 
> use the new flex, because if they don't the build will fail anyway, at any of 
> the other lexers.

It's been true for quite a long time that only ecpg's lexer generated
any warnings on stable flex releases like 2.5.4.  The Makefile
configuration was not unreasonable.
        regards, tom lane

On Tue, Mar 18, 2008 at 07:23:40PM +0100, Peter Eisentraut wrote:
> Am Dienstag, 18. März 2008 schrieb Bruce Momjian:
> > Peter Eisentraut wrote:
> > > Log Message:
> > > -----------
> > > Don't need -Wno-error anymore, because flex is no longer producing
> > > warnings.
> >
> > I see this patch only affects ecpg?  Is this related to Tom's report
> > from trying the new flex?
> 
> Yes.  There used to be a few -Wno-error uses throughout the source code, near 
> the flex calls, to work around the well-known flex warnings issue in case 
> someone wanted to use -Werror to check his builds.  The fact that there was 
> only one such use left in the ecpg subtree shows that this wasn't really 
> maintained and could never have worked reliably.

There currently are a few warnings (with gcc 4.3.0):
bootscanner.c:1461: warning: 'input' defined but not used
scan.c:6670: warning: 'input' defined but not used
guc-file.c:1233: warning: 'input' defined but not used
tuplesort.c:2475: warning: inlining failed in call to 'myFunctionCall2': --param large-stack-frame-growth limit
reached
tuplesort.c:2526: warning: called from here
tuplesort.c:2475: warning: inlining failed in call to 'myFunctionCall2': --param large-stack-frame-growth limit
reached
tuplesort.c:2526: warning: called from here
[...]
tqual.c: In function 'HeapTupleSatisfiesVacuum':
tqual.c:115: warning: inlining failed in call to 'SetHintBits': call is
unlikely and code size would grow
tqual.c:1084: warning: called from here
tqual.c:115: warning: inlining failed in call to 'SetHintBits': call is
unlikely and code size would grow
tqual.c:1088: warning: called from here
[...]
pgc.c:3507: warning: 'input' defined but not used
psqlscan.c:3853: warning: 'input' defined but not used
pl_scan.c:2458: warning: 'input' defined but not used


Kurt