On 04/07/2016 12:44, Aki Tuomi wrote:
On 04.07.2016 14:24, Juan C. Blanco wrote:
On 01/07/2016 18:51, Timo Sirainen wrote:
On 01 Jul 2016, at 19:09, Juan C. Blanco <jcblanco@fi.upm.es> wrote:
I Haven't had the time to check the sha1.h error with the new fixes but I've just done so after de 2.2.25 release was out and I'm having the same error:
gcc -DHAVE_CONFIG_H -I. -I. -I../.. -std=gnu99 -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-cast -fno-builtin-strftime -Wstrict-aliasing=2 -I/usr/kerberos/include -c guid.c -fPIC -DPIC -o .libs/guid.o In file included from guid.c:6: sha1.h:80: error: static or type qualifiers in abstract declarator
I thought this was fixed in 2.2.25.. In config.h you should have only:
#define STATIC_ARRAY
(not #define STATIC_ARRAY static)
OK, my config.h contains:
/* C99 static array */ #define STATIC_ARRAY static
Regards.
Operating System: CentOS 5.11 GCC Version: gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-55)
We have four systems like this with dovecot 2.2.24 working fine. I would like to know if this platform will not be compatible anymore with dovecot and if we need to upgrade our systems prior to dovecot version since, besides this, I have the problem related to the openssl version for the lib-dcrypt library
I recommend upgrading. We're considering not supporting older OSes anymore at all..
Hi!
This problem is hopefully now fixed by https://github.com/dovecot/core/commit/d9c865ce774aae9f2f17b89e7e94c3cfca29d...
Aki, I had a go mock building on CentOS5 and it appears that the test succeeds with this patch in place:
configure:21182: checking if we can use C99 static in array sizes configure:21202: gcc -c -std=gnu99 -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -fno-strict-aliasing -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-cast -fno-builtin-strftime -Wstrict-aliasing=2 conftest.c >&5 configure:21202: $? = 0 configure:21214: result: yes
(confirmed was built using unsigned char array)
In fact, after a cursory examination and some testing I suspect the CentOS5 gcc does nominally support static array sizes, and it's likely hitting a compiler bug on the build proper?
All I can suggest is a manual configure flag toggle of the test, or that this bug is better placed against Redhat? I haven't been able to narrow down a reproducer on triggering the error message other than during build.
-- Dave