Control: clone 751054 -2 Control: reassign -2 pidgin-sipe 1.18.1-1 Control: severity -2 serious Control: retitle -2 pidgin-sip: FTBFS - uses -Werror -Wundef with third-party headers
> Michael Tautschnig <m...@debian.org> writes:
>
> > libtool: compile: gcc -DHAVE_CONFIG_H -I. -I../.. -D_FORTIFY_SOURCE=2
> > -Werror -Wall -Wextra -Waggregate-return -Wcast-align
> > -Wdeclaration-after-statement -Wdeprecated-declarations -Winit-self
> > -Wmaybe-uninitialized -Wmissing-declarations -Wmissing-prototypes
> > -Wnested-externs -Wpointer-arith -Wundef -Wunused-but-set-variable
> > -I/usr/include/glib-2.0 -I/usr/lib/x86_64-linux-gnu/glib-2.0/include
> > -DLOCALEDIR=\"/usr/share/locale\" -I./../api -I/usr/include/mit-krb5 -g -O2
> > -fstack-protector --param=ssp-buffer-size=4 -Wformat
> > -Werror=format-security -c sip-sec-gssapi.c -fPIC -DPIC -o
> > .libs/libsipe_core_la-sip-sec-gssapi.o
> > In file included from sip-sec-gssapi.c:42:0:
> > /usr/include/mit-krb5/gssapi/gssapi.h:47:5: error: "TARGET_OS_MAC" is not
> > defined [-Werror=undef]
> > #if TARGET_OS_MAC
> > ^
>
> -Werror=undef is not really a sane compiler flag to try to build with if
> you use random third-party headers. Most software does not have -Wundef
> cleanliness as an explicit goal, and the machinations that you have to go
> through to achieve this are often not trivial.
>
I tend to agree, hence cloned the bug to get pidgin-sipe to change (as well).
> In this specific case, it would probably be fine for the MIT Kerberos code
> base to change this line to instead be:
>
> #if !defined(TARGET_OS_MAC)
^ (Sure? I don't think negation is due here.)
>
> but it would surprise me if this is the last problem that you run into
> along these lines. I would instead remove either -Werror or -Wundef from
> your compiler flags or otherwise make -Wundef warnings not a fatal error.
>
So I do believe that
^\s*#\s*if\s+[a-zA-Z_][a-zA-Z0-9_]*\s*$
is almost always wrong as those should really be #ifdef ... or #if defined(...),
but there may be more complex expressions that are wrong as well. Fixing this
appears to be a worthwhile (but indeed possibly non-trivial) goal.
Best,
Michael
pgpgs6VDMPyIS.pgp
Description: PGP signature
