Mailing List Archive

Date: Tue, 29 Aug 95 17:14 BST
From: drtr@ast.cam.ac.uk (David Robinson)

** Nearly there, but it needs to check the return value from
gethostbyaddr(), otherwise Apache could core dump. (As in the
bug fixed by bad-namesrvr.)

Sigh... this smacks of the time that you vetoed the original DBM auth
patches because some function wasn't prototyped the way you'd have
liked. Are you quite sure that a veto is an appropriate way to
register this sort of objection?

rst

> I may well be missing something here, but if my reading of this
> paragraph is correct, Randy's approach can be made to work as follows:
>
> #ifdef BSD_4_4
> #if BSD_4_4 > 0.5
> sprintf(tag,"%qd",finfo.st_size);
> #else
> sprintf(tag,"%ld",finfo.st_size);
> #endif /* *real* BSD_4_4 */
> #else
> sprintf(tag,"%ld",finfo.st_size);
> #endif
>
> This will, of course, win no prizes for elegance or concision;
> however, if it will in fact work on all BSD-oid systems, I'd much
> prefer it. The advantage is that it doesn't involve messing around
> much with the rest of the code, including the config defines for
> systems which we don't necessarily have (and which I *really* don't
> want to mess with until post-1.0, at which point conf.h clearly does
> deserve serious rework).
>
> rst

That won't work, but the following will. The preprocessor
complains about floating point numbers in #if with the above
directives. Anyone know *who* decided to put that in sys/param.h?

#if BSD
#if BSD > 199305
sprintf(tag,"%qd",finfo.st_size);
#else
sprintf(tag,"%ld",finfo.st_size);
#endif /* *real* BSD4_4 */
#else
sprintf(tag,"%ld",finfo.st_size);
#endif

I'll update the patch in the morning if that's OK.

My votes: '0' means I did not test the specific problem it addresses

+1 patch11.access_cleanup
+1 patch11.bad-namesrvr
+1 patch11.bsdi-rfc931
-1 patch11.bsdi-longsz
+1 patch11.cgi-cleanup
+1 patch11.config-script
0 patch11.dyn-load
+1 patch11.htaccess-first
+1 patch11.imap_point
+1 patch11.inc-cgi-pause
+1 patch11.include-opt
+1 patch11.include-text
0 patch11.lookup_uri
+1 patch11.no-auth-no-error
+1 patch11.no-bsd-conf rev3
0 patch11.rfc931
+1 patch11.virt-srvr-name

X-Uri: http://www.zyzzyva.com/
Date: Tue, 29 Aug 1995 23:31:28 -0500
From: Randy Terbush <randy@zyzzyva.com>

That won't work, but the following will. The preprocessor
complains about floating point numbers in #if with the above
directives. Anyone know *who* decided to put that in sys/param.h?

#if BSD
#if BSD > 199305
sprintf(tag,"%qd",finfo.st_size);
#else
sprintf(tag,"%ld",finfo.st_size);
#endif /* *real* BSD4_4 */
#else
sprintf(tag,"%ld",finfo.st_size);
#endif

Sigh... this won't work *either* --- it'll generate a parse error on
systems where BSD is not defined right at the first #if, which should
be a #ifdef. I am teetering on changing my vote on bsd-no-conf to a
veto after all these changes, just because this round of continual
changes has left me unsure what people are voting on, or whether they
would approve of the current version.

rst

+1 patch11.htaccess-first --- Lets .htaccess override <Directory> in
access control, etc. [David Robinson]

I tested this thoroughly, but it made no difference on my system
since I don't have any brain-damaged access configurations. This
is only a bug fix in that it retains compatibility with NCSA.

0 patch11.imap_point --- add "point" directive to imagemap module [James
Cloos]

I would vote +1 on this, but I don't have Imaps on my server.
However, I disagreee with David in that this is more a bug fix
then was the patch11.htaccess-first, since this retains compatibility
for people with non-broken configurations.

+1 patch11.include-opt --- properly honor Options Includes (or more
accurately, properly honor turning it off). [David Robinson]

Tested.

+1 patch11.include-text --- Make IncludesNoExec work properly, by being
willing to include at least text/*. [Roy Fielding]

Tested.

-1 patch11.no-bsd-conf --- Delete BSD #define from conf.h everyplace; also
cures the bsdi-longsz problem above. [ Randy Terbush ]

Does not work as advertized.

0 on the rest, since our planning meetings here prevented any
further testing on my part.

......Roy

Rst wrote:
> Date: Tue, 29 Aug 95 17:14 BST
> From: drtr@ast.cam.ac.uk (David Robinson)
>
> ** Nearly there, but it needs to check the return value from
> gethostbyaddr(), otherwise Apache could core dump. (As in the
> bug fixed by bad-namesrvr.)
>
>Sigh... this smacks of the time that you vetoed the original DBM auth
>patches because some function wasn't prototyped the way you'd have
>liked. Are you quite sure that a veto is an appropriate way to
>register this sort of objection?

Well, it does introduce a bug that (IMHO) would need fixing, and the
bug it does fix is not terribly urgent, so I was working on the principle
that a single correct patch was better than a buggy patch followed by
another fix.

David.

Rst wrote:
> Date: Tue, 29 Aug 95 17:14 BST
> From: drtr@ast.cam.ac.uk (David Robinson)
>
> ** Nearly there, but it needs to check the return value from
> gethostbyaddr(), otherwise Apache could core dump. (As in the
> bug fixed by bad-namesrvr.)
>
>Sigh... this smacks of the time that you vetoed the original DBM auth
>patches because some function wasn't prototyped the way you'd have
>liked. Are you quite sure that a veto is an appropriate way to
>register this sort of objection?

Well, it does introduce a bug that (IMHO) would need fixing, and the
bug it does fix is not terribly urgent, so I was working on the principle
that a single correct patch was better than a buggy patch followed by
another fix.

David.