Re: [HACKERS] PL compile warning messages

Bruce wrote:
>
> >     I get them too and don't know exactly what they mean. I think
> >     gcc is just  telling  that  after  returning  to  the  setjmp
> >     location  via  longjmp  the  variable  might not contain what
> >     there was before (the longjmp  mechanism  only  restores  the
> >     stack pointers, not the variable contents on the stack).
> >
> >     But  the  longjmp's  are  there  only  to  clean up the Tcl's
> >     interpreter  nesting  and  allocations  in  the  case  of  an
> >     elog(ERROR) before really jumping back into the postgres main
> >     loop. Tcl doesn't allocate via  palloc,  so  there  would  be
> >     memory  allocations  never  free'd  otherwise.  The mentioned
> >     variables aren't accessed after the longjumping session began
> >     (it's  really a longjmp party if Tcl triggers use in turn Tcl
> >     functions where the queries invoke  other  functions/triggers
> >     and so on :-).
>
> See postgres.c.  We used to have that warning in postgres.c, but someone
> changed something to fix it.  I can't see what was changed, now that I
> am looking at it.

    I took a lood at it and didn't saw the changes either. Then I
    played around with the code.

    In some cases only a strange workaround  could  prevent  that
    warning.  Creating  another  variable  of  the  same type and
    somewhere in the function doing var2 = var1; and  then  using
    var2  instead  (doesn't do anything useful and makes the code
    more obscure).

    From the gcc manpage:

       -W     Print extra warning messages for these events:

          ·   A nonvolatile automatic variable might  be  changed
              by  a call to longjmp.  These warnings are possible
              only in optimizing compilation.

              The compiler sees only the  calls  to  setjmp.   It
              cannot  know where longjmp will be called; in fact,
              a signal handler could call it at any point in  the
              code.  As a result, you may get a warning even when
              there is in fact no problem because longjmp  cannot
              in  fact be called at the place which would cause a
              problem.

    In fact I think it's legal to ignore these  warnings  because
    there is in fact no place which would cause a problem.

    And in fact I love this snippet of the manpage :-)


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#======================================== jwieck@debis.com (Jan Wieck) #