| Message ID | 20240411143025.352507-8-pvorel@suse.cz |
|---|---|
| State | New |
| Headers | show |
| Series | Remove UCLINUX (nommu support) from LTP legacy C API | expand |
Hi! > +#ifndef __UCLIBC__ The UCLIBC should be removed as well, since that was libc for uCLinux. > #include <aio.h> > #endif > #endif > -- > 2.43.0 > > > -- > Mailing list info: https://lists.linux.it/listinfo/ltp
Hi Cyril, > Hi! > > +#ifndef __UCLIBC__ > The UCLIBC should be removed as well, since that was libc for uCLinux. IMHO you are wrong here. IMHO uCLinux-ng [1], still used libc (used in Buildroot and openembedded). And log in e5e938f4b7 (2009) suggests that it's about libc. But the question is whether it's still needed. I could try to remove these and test, but as a separate effort. We have few of them: $ git grep -l __UCLIBC__ include/old/tlibio.h lib/tlibio.c testcases/kernel/syscalls/fmtmsg/fmtmsg01.c utils/benchmark/ebizzy-0.3/ebizzy.h Kind regards, Petr [1] https://cgit.uclibc-ng.org/cgi/cgit/uclibc-ng.git/tree/include/features.h#n185
Hi! > IMHO you are wrong here. IMHO uCLinux-ng [1], still used libc (used in Buildroot > and openembedded). And log in e5e938f4b7 (2009) suggests that it's about libc. > But the question is whether it's still needed. I could try to remove these and > test, but as a separate effort. You are right, these should be converted into configure checks later on. And the tlibio should probably be rewritten completely. So let's keep the patch as it is.
Hi! On 4/12/24 09:36, Cyril Hrubis wrote: > Hi! >> IMHO you are wrong here. IMHO uCLinux-ng [1], still used libc (used in Buildroot >> and openembedded). And log in e5e938f4b7 (2009) suggests that it's about libc. >> But the question is whether it's still needed. I could try to remove these and >> test, but as a separate effort. > You are right, these should be converted into configure checks later on. > > And the tlibio should probably be rewritten completely. So let's keep > the patch as it is. > I started to rewrite part of that library when I was working on growfiles test rewrite, but the complexity of it, the amount of nested statements and macro directives made me crazy. Waiting for a better moment to restart rewriting it, but work is on-going. Andrea
Hi! > >> IMHO you are wrong here. IMHO uCLinux-ng [1], still used libc (used in Buildroot > >> and openembedded). And log in e5e938f4b7 (2009) suggests that it's about libc. > >> But the question is whether it's still needed. I could try to remove these and > >> test, but as a separate effort. > > You are right, these should be converted into configure checks later on. > > > > And the tlibio should probably be rewritten completely. So let's keep > > the patch as it is. > > > I started to rewrite part of that library when I was working on > growfiles test rewrite, but the complexity of it, the amount of nested > statements and macro directives made me crazy. > Waiting for a better moment to restart rewriting it, but work is on-going. Maybe it would be better to do this in a steps, first remove everything that is inside the ifdefs for hpux, sgi etc. and only then start working with the actual code...
Hi! On 4/12/24 10:57, Cyril Hrubis wrote: > Hi! >>>> IMHO you are wrong here. IMHO uCLinux-ng [1], still used libc (used in Buildroot >>>> and openembedded). And log in e5e938f4b7 (2009) suggests that it's about libc. >>>> But the question is whether it's still needed. I could try to remove these and >>>> test, but as a separate effort. >>> You are right, these should be converted into configure checks later on. >>> >>> And the tlibio should probably be rewritten completely. So let's keep >>> the patch as it is. >>> >> I started to rewrite part of that library when I was working on >> growfiles test rewrite, but the complexity of it, the amount of nested >> statements and macro directives made me crazy. >> Waiting for a better moment to restart rewriting it, but work is on-going. > Maybe it would be better to do this in a steps, first remove everything > that is inside the ifdefs for hpux, sgi etc. and only then start working > with the actual code... > Well that's the easiest part. growfiles has a really huge amount of bad code, it's the biggest in the whole LTP suite and (unfortunately) it's a really important one. It will take a bit...probably it will be incremental. Andrea
diff --git a/lib/tlibio.c b/lib/tlibio.c index bfb4ea023..ed82a85b4 100644 --- a/lib/tlibio.c +++ b/lib/tlibio.c @@ -108,7 +108,7 @@ #include <string.h> #endif #if defined(__linux__) || defined(__sun) || defined(__hpux) || defined(_AIX) -#if !defined(UCLINUX) && !defined(__UCLIBC__) +#ifndef __UCLIBC__ #include <aio.h> #endif #endif
Signed-off-by: Petr Vorel <pvorel@suse.cz> --- lib/tlibio.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)