autoconf ignoring CFLAGS? - c

Question

I'm building a package (vice 3.4) that uses GNU autoconf. Running ./configure fails with the following message:
checking size of time_t... 0
configure: error: can not figure type of time_t
error: Bad exit status from /var/tmp/rpm-tmp.wIgnPw (%build)
Examining config.log, this appears to be failing due to the following error:
/usr/bin/ld: /tmp/ccMTSdtB.o: relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/ld: final link failed: nonrepresentable section on output
collect2: error: ld returned 1 exit status
I tried resolving this by adding -fPIC to the CFLAGS environment variable:
CFLAGS=-fPIC ./configure
But while this is clearly being used during other stages of ./configure...
configure:3657: checking whether the C compiler works
configure:3679: gcc -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -D_GNU_SOURCE=1 -fPIC -Wl,-z,relro -Wl,--as-needed -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld conftest.c >&5
It doesn't seem to be used in the feature tests related to time_t:
configure:9684: checking for time_t in time.h
configure:9700: gcc -c -g -O3 -Wall -Wformat -Wformat-signedness -Wshadow -Wpointer-arith -Wstrict-prototypes -Wuninitialized -Wunreachable-code -Wno-unused-parameter -Werror=implicit-function-declaration -Wfatal-errors conftest.c >&5
[...]
configure:9753: checking size of time_t
configure:9758: gcc -o conftest -g -O3 -Wall -Wformat -Wformat-signedness -Wshadow -Wpointer-arith -Wstrict-prototypes -Wuninitialized -Wunreachable-code -Wno-unused-parameter -Werror=implicit-function-declaration -Wfatal-errors -Wl,-z,relro -Wl,--as-needed -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld conftest.c >&5
Looking at configure.ac, the above code is generated from:
if test $bu_cv_decl_time_t_time_h = yes; then
AC_CHECK_SIZEOF([time_t],[],[#include <time.h>])
else
AC_CHECK_SIZEOF([time_t],[],[#include <sys/types.h>])
fi
I've been able to work around the problem by redefining CC instead:
CC="gcc -fPIC" ./configure
This works, but it's ugly (what if something wants to call gcc without -fPIC?). Is there a reason that AC_CHECK_SIZEOF is ignoring CFLAGS?

Answer 1

It looks like the suspicion I expressed in comments is confirmed. The VICE configure program intentionally substitutes its own choice of CFLAGS for the user's specified ones when it performs some of its checks, including the ones related to time_t. Thus, this is not a question of Autoconf ignoring your flags, but rather of VICE doing so.
I reiterate that I think it's a more significant issue that configuration doesn't work out of the box for you in the first place. I suggest raising that as an issue against the project. Although I don't like that VICE is so cavalier about ignoring the user's CFLAGS, that's a separate matter, and one that I suspect wouldn't get much traction with the devs.

Answer 2

John correctly identified the issue, and based on that information I did a little digging and I've figure out what's going on.
I'm building this under the control of rpm, which calls configure like this:
%configure --enable-native-gtk3ui $COMMON_FLAGS
That %configure expands into a chunk of shell code that looks like this:
CFLAGS="${CFLAGS:--O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection}" ; export CFLAGS ;
CXXFLAGS="${CXXFLAGS:--O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection}" ; export CXXFLAGS ;
FFLAGS="${FFLAGS:--O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -I/usr/lib64/gfortran/modules}" ; export FFLAGS ;
FCFLAGS="${FCFLAGS:--O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -I/usr/lib64/gfortran/modules}" ; export FCFLAGS ;
LDFLAGS="${LDFLAGS:--Wl,-z,relro -Wl,--as-needed -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld}" ; export LDFLAGS;
[ "1" = 1 ] && for i in $(find $(dirname ./configure) -name config.guess -o -name config.sub) ; do
[ -f /usr/lib/rpm/redhat/$(basename $i) ] && /usr/bin/rm -f $i && /usr/bin/cp -fv /usr/lib/rpm/redhat/$(basename $i) $i ;
done ;
[ "1" = 1 ] && [ x != "x-Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld" ] &&
for i in $(find . -name ltmain.sh) ; do
/usr/bin/sed -i.backup -e 's~compiler_flags=$~compiler_flags="-Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld"~' $i
done ;
./configure --build=x86_64-redhat-linux-gnu --host=x86_64-redhat-linux-gnu \
--program-prefix= \
--disable-dependency-tracking \
\
--prefix=/usr \
--exec-prefix=/usr \
--bindir=/usr/bin \
--sbindir=/usr/sbin \
--sysconfdir=/etc \
--datadir=/usr/share \
--includedir=/usr/include \
--libdir=/usr/lib64 \
--libexecdir=/usr/libexec \
--localstatedir=/var \
--sharedstatedir=/var/lib \
--mandir=/usr/share/man \
--infodir=/usr/share/info
Of particular interest is the fact that this is setting LDFLAGS. The failing config test is:
gcc -o conftest -g -O3 -Wall -Wformat -Wformat-signedness -Wshadow \
-Wpointer-arith -Wstrict-prototypes -Wuninitialized -Wunreachable-code \
-Wno-unused-parameter -Werror=implicit-function-declaration \
-Wfatal-errors -Wl,-z,relro -Wl,--as-needed -Wl,-z,now \
-specs=/usr/lib/rpm/redhat/redhat-hardened-ld conftest.c
That -specs line points to a gcc spec file that looks like this:
*self_spec:
+ %{!static:%{!shared:%{!r:-pie}}}
That effectively adds the -pie option to the command line. We can reproduce the failure with a simple hello.c program:
#include <stdio.h>
#include <unistd.h>
int main(int argc, char **argv) {
printf("hello world\n");
return(0);
}
Like this:
$ gcc -o hello -pie hello.c
/usr/bin/ld: /tmp/ccXRAu6r.o: relocation R_X86_64_32 against `.rodata' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/ld: final link failed: nonrepresentable section on output
collect2: error: ld returned 1 exit status
But here's the thing:
A successful gcc invocation looks like this:
gcc -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 \
-Wp,-D_GLIBCXX_ASSERTIONS -fexceptions -fstack-protector-strong \
-grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 \
-specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic \
-fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection \
-D_GNU_SOURCE=1 -fPIC -Wl,-z,relro -Wl,--as-needed -Wl,-z,now \
-specs=/usr/lib/rpm/redhat/redhat-hardened-ld conftest.c
Note that there is a second -specs option there, that points to /usr/lib/rpm/redhat/redhat-hardened-cc1, which looks like:
*cc1_options:
+ %{!r:%{!fpie:%{!fPIE:%{!fpic:%{!fPIC:%{!fno-pic:-fPIE}}}}}}
This takes care of adding -fPIE to the command line, which results in a successful compile:
$ gcc -o hello -pie -fPIE hello.c
$
We're hitting this failure because the configure script is replacing CFLAGS during the tests, but preserving LDFLAGS. This results in a faulty command line.
For my own purposes, I'm just patching the configure script to not replace CFLAGS during the time_t related tests. This seems both expedient (it's a one-character diff if I just comment out the appropriate line) and effective (it results in a successful build).
I will probably report this upstream as a bug.