Re: rcs-5.8 test failures on HP-UX 11.31

() Paul Ackersviller <[email protected]>
() Mon, 5 Mar 2012 00:31:08 +0000

   slightly newer piece of hardware now, perhaps with some different
   vendor patches, although my OS version is still the same.

OK, fingers crossed...

   The good and bad news is that I'm not even able to get a
   successfult compile now.  I expect at least that things should be
   easier to recify then.

Would you be willing to test a version from the public Git repository?

   Both different gcc version I have here end with
   gcc -std=gnu99 -std=gnu99 -DHAVE_CONFIG_H -I. -I../lib  -I'../lib'    -g -MT ci.o -MD -MP -MF .deps/ci.Tpo -c -o ci.o ci.c
   ci.c: In function 'main':
   ci.c:678:15: error: 'help' undeclared (first use in this function)
   ci.c:678:15: note: each undeclared identifier is reported only once for each function it appears in
   ci.c:682:3: warning: implicit declaration of function 'CHECK_HV' [-Wimplicit-function-declaration]

That seems to indicate a problem with the preprocessor, which is
involved in defining each ‘help’ (via #include "ci.help"), which in
turn defines ‘CHECK_HV’ (via #include "gnu-h-v.h").  Or perhaps the
fault lies w/ the program that produces ci.help.

   However, since I'm not entirely sure how well gcc works on this
   platform yet, I'm more inclined to use HP's compiler... but it
   ends similarly with

           source='ci.c' object='ci.o' libtool=no \
           DEPDIR=.deps depmode=hp2 /bin/sh ../build-aux/depcomp \
           cc -DHAVE_CONFIG_H -I. -I../lib  -I'../lib'    -AC99 -c ci.c
   "ci.c", line 249: warning #2186-D: pointless comparison of unsigned integer with zero
               if (!removedlock && 0 <= (removedlock = removelock (targetdelta)))
                                     ^

Good catch.  The variable ‘removedlock’ used to be ‘int’, but i
changed it to ‘bool’ without considering that the return value of
‘removelock’ is int, and may indeed be negative.  The assignment
loses information.  This is now fixed with:

 http://git.savannah.gnu.org/cgit/rcs.git/commit/?id=8540432c56f3

Someone needs to perform an audit for this kind of mistake...

   "ci.c", line 678: error #2020: identifier "help" is undefined
           .help = help,
                   ^

   "ci.c", line 682: warning #2223-D: function "CHECK_HV" declared implicitly
       CHECK_HV ();
       ^

Could you please try (in src/ subdir)

 rm -f ci.help && make ci.help

and let me know if the output differs (substantially) from the
sample from my computer, here attached?