login
Search
Hi all, you might know that since ~ 2 years, the Sun Studio compilers are available for Linux. Given the fact that they typically produce faster code than GCC and that they offer more debug/optimizing features, they are worth testing. While it is no problem to use Sun Studio for non-Linux-specific programs, it is impossible to compile programs using Sun Studio if these programs offer Linux specific features. Star ftp://ftp.berlios.de/pub/star/alpha/ offers support for archiving ext2 file flags. If star is compiled using Sun Studio, star's autoconfiguration will disable support for the Linux specific features because it detects "broken" include files. Cdrtools ftp://ftp.berlios.de/pub/cdrecord/alpha/ offer support for an OS dependent SCSI transport. Cdrtools cannot be compiled wihout support for SCSI transport, so it is impossible to use Sun Studio to compile cdrtools. Why does this happen? Well, the reason is that in order to support Linux specific features, you need to include Linux specific include files (the Linux kernel include files). As these include files are currently not written in vanilla (ANSI) C but in a GCC-C-variant, other compilers do not like these include files. Is there some hope that at least the Linux kernel interface definition files and everything recursively included from these files will be rewritten in vanilla ANSI C? Jörg -- EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin js@cs.tu-berlin.de (uni) schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily -
I assume you typoed and meant "cleaned up kernel include files as can you give a specific example of a header installed by make headers_install that breaks this way and is hurting you? Because it may well be possible to fix the problems, now that we have this special cleanup phase since several releases.... -
I am thinking about kernel include files that do correct preincludes for
star needs "ext2_fs.h". This file is not usable at all on many Linux
distributions, even with GCC.
libscg (cdrtools) needs "scsi/sg.h" but it currently includes a lot of other
files:
scsi-linux-sg.c:#include <linux/version.h>
scsi-linux-sg.c:#include <asm/types.h>
scsi-linux-sg.c:#include <scsi/scsi.h>
scsi-linux-sg.c:#include <linux/scsi.h>
scsi-linux-sg.c:#include <linux/fs.h> /* From ancient versions, really needed? */
scsi-linux-sg.c:#include "block/blk.h" /* From ancient versions, really needed? */
scsi-linux-sg.c:#include "scsi/scsi.h"
scsi-linux-sg.c:#include "scsi/sg.h"
scsi-linux-sg.c:#include <linux/cdrom.h>
If there wase _one_ clean SCSI pass through interface on Linux,
things would be a lot easier.....
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
there is no __KERNEL__ in the "make headers_install"'d kernel headers. And they are more or less sanity checked (no sanity check is complete obviously) against certain common errors already (like including headers that aren't installed). if sanity checks are missing (and you seem to imply they do) then I'm oh I thought we were talking about the kernel, not about distributions. Again, did you check the output of "make headers_install" (this is what distributions are using going forward), and if so, can you give *specific* examples of what we should add checks and fixes for? -- if you want to mail me at work (you don't), use arjan (at) linux.intel.com Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org -
not quite technically true, as "unifdef" isn't smart enough to factor out __KERNEL__ if it's part of a larger, logical expression involving the "||" operator. but that's being admittedly nitpicky. rday -- ======================================================================== Robert P. J. Day Linux Consulting, Training and Annoying Kernel Pedantry Waterloo, Ontario, CANADA http://fsdev.net/wiki/index.php?title=Main_Page ======================================================================== -
I was curious so I did:
$ mkdir ~/foo
$ cd ~/kernel/linux-2.6
$ make INSTALL_HDR_PATH=~/foo
$ cd ~/foo
$ cat j.c
#include <stdio.h>
#include "etx2_fs.h"
main()
{
printf("helloo\n");
}
$ gcc -o j j.c
=> No warning, no errors
$ gcc -ansi -pedantic -o j j.c
In file included from j.c:2:
ext2_fs.h:53:25: warning: ISO C does not permit named variadic macros
$ gcc -ansi -pedantic -std=c99 -o j j.c
In file included from j.c:2:
ext2_fs.h:53:25: warning: ISO C does not permit named variadic macros
j.c:5: warning: return type defaults to ‘int’
Is it this part you are referring to in the above or have I overlooked something?
Sam
-
this has been discussed many times and the answer is that the kernel is not gong to change it's side of things to ANSI C. that doesn't mean that one of the many projects out there to create seperate interface headers won't do this. however, there is another interesting thing going on at the moment. The standards commity is currently deciding what will be in the next rev of the C specs (tentitivly labels C1x for discussions, they are hopeing to release them in '09 or '10) some of the things that are GCC extentions are going to be added to the spec (ahtough possibly with a change to the syntax) so now is the time to talk to your local reps who go to these meetings and make your suggestions for what should be added to the standard and what should not be. remember that anything to be added must be implemented somehwere, preferrably by multiple seperate compilers. David Lang -
I don't think that's entirely true with regard to the include files. We have always tried not to step on anyone's toes there, e.g. regarding the use of __u32 vs. uint32_t style types. It's certainly desirable to make the kernel headers that are _meant_ for inclusion compatible with standard compilers. Mike Frysinger has posted a few patches that make the installed headers friendlier to strict C99 users. While there was some negative feedback about these patches, it was not about the idea of making the installed headers C99 clean, but rather about the question whether those non-clean parts should be exported in the first place. Now whether a specific header file should be installed and potentially included in user space is certainly debatable in many cases, but at least it's now clearly defined through the include/*/Kbuild files. If someone has a good reason to change which files are exported, he Those projects are all practically dead, since we have the 'make headers_install' target in the Linux source. Arnd <>< -
Wouldn't it be simpler to ask the developers to deliver their include files
in a state that is clean for user space programs?
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
The main problems are not really hard to fix......
- Most problems eem to be related to the fact that Linux does not
use C-99 based types in the kernel and the related type definitions
are not written in plain C. This is something that should be fixed
with a source consolidation program or by defining aliases to
C-99 types in case the compiler is not GCC.
- Other problems are caused by additional tag definitions that could
be disabled in case of a non-GCC compile.
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
The argument has been made that the standard C99 types are _optional_, and anything included from a C library's headers without _explicitly_ being included by the user shouldn't define those types. Personally, I think that's a load of bollocks. And it certainly doesn't apply to Linux-specific files like <linux/cdrom.h>, which are perfectly entitled to use a C standard from last millennium, regardless of namespace 'pollution' issues. That's why we continue to use the crappy __u32 types. Can you be more specific about why this is a problem? Don't we mostly define those crappy types using arch-specific knowledge, as We mostly try to remove this from user-visible parts of exported headers. Sometimes we just remove it altogether; other bits are stripped at export time when you run 'make headers_install'. Again, can you be more specific about the problem? -- dwmw2 -
It would certainly help if Joerg would tell what exactly breaks, but I
spot one likely problem in include/asm-i386/types.h:
#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
typedef __signed__ long long __s64;
typedef unsigned long long __u64;
#endif
It might make sense to remove the #if and simply require that
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
-
Gah, this particular topic and a few other similar header- compatibility ones show up once a month on LKML; I should probably just make a patch to fix all the types.h files and be done with it. The proper solution is this: # if __STDC_VERSION__ >= 19901L typedef signed long long __s64; typedef unsigned long long __u64; # elif defined(__GNUC__) __extension__ typedef signed long long __s64; __extension__ typedef unsigned long long __u64; # else # error "Your compiler doesn't support long long (IOW: It sucks). Please get a new one" # endif That way if you have any kind of vaguely-long-long-compatible compiler then it will work, and otherwise you'll get a nice useful error message. It also makes sure that GCC doesn't spew warnings/ errors when in c89-pedantic mode. The "__extension__" keyword is designed for use in implementation header files which want to use GCC- isms unconditionally. Cheers, Kyle Moffett -
This part is wrong.
gcc supported "long long" before C99 existed.
That's a good point I missed.
What about:
#if defined(__GNUC__) && __STDC_VERSION__ < 19901L
__extension__ typedef signed long long __s64;
__extension__ typedef unsigned long long __u64;
#else
typedef signed long long __s64;
typedef unsigned long long __u64;
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
-
What about using:
#if (defined(__GNUC__) || defined(__SUNPRO_C)) && !defined(__STRICT_ANSI__)
Well, there seems to be one other compiler (the one from Intel).
It may be that if this compiler does not claim to be GCC, another
definituion needs to be added.
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
AFAIR __extension__ was gcc specific.
Does the Sun cc provide options for strict C90 checking?
And if yes, what is it's syntax for disabling this checking for a line
AFAIR the Intel compiler claims to be gcc.
But these are by far not the only C compilers under Linux, and the more
important points are:
Is there any userspace Linux compiler that does not support "long long"?
If yes, is there any other way to tell that something is a
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
-
TenDRA C: "test.c", line 6: Error: [ISO 6.5.2]: Illegal type specifier, 'long long', assuming 'long'. -- Hilsen Harald. -
You cannot use this compiler for any program that uses an interface that
needs long long.
BTW: C99 requires long long but the Large File summit from 1995 did to the
same since 1995. An 32 bit OS that supports large files cannot have a compiler
that does not support long long.
If TenDRA C does support 64 bit integral type in a different way. TenDRA C
would need to deliver include files that take care about this fact.
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
I can't even find the docs for their "tcc". Their "tchk" appears to have a "#pragma longlong type allow" or something, so I'd imagine the same exists for tcc and would be required to build stuff using kernel headers. On the other hand, their compiler looks so immature that it does not appear to be worth spending much time worrying about now. When somebody shows up with a solution for that compiler then we can look at it at that time. Cheers, Kyle Moffett -
tchk is the same thing pretty much with output disabled. There's a HTML copy of the man page here: http://www.penguin-soft.com/penguin/man/1/tendracc.html It's packaged for Debian and therefore Ubuntu and other derived You just need to specify -Ysystem (or some other API selection option) when building to get it to accept long long. Since TenDRA focuses on strict standards conformance it defaults to something roughly equivalent to GCC with -std=3Dc89 -pedantic -Werror and requires the user to explicitly enable support for any other APIs and features they want to The compiler is solid enough but old - it predates C99 and has had no real development since then beyond updating the system include overrides to work with newer glibc versions. --=20 "You grabbed my hand and we fell into it, like a daydream - or a fever."
Not that I know of. Probably the straight #else conditional is OK. We should also merge up the types since *EVERY* linux architecture has these same types: typedef signed char __s8; typedef unsigned char __u8; typedef signed short __s16; typedef unsigned short __u16; typedef signed int __s32; typedef unsigned int __u32; Then all 64-bit archs have: typedef signed long __s64; typedef unsigned long __u64; While all 32-bit archs have: typedef signed long long __s64; typedef unsigned long long __u64; The only trick is if you care about building 32-bit compat code using 64-bit linux kernel headers. In that case we should probably just make all archs use "long long" for their 64-bit integers, unless there's some platform I'm not remembering where "long long" is 128- bits or bigger. The other benefit is that people could then just use the printf format "%llu" for 64-bit integers instead of having to conditionalize it all over the place. I'm working on a patch now. Cheers, Kyle Moffett -
LDD3 ch. 11 says that long on Sparc64 is 32 bits. Same for "ppc" (don't know which power* arch. they mean by that). --- ~Randy *** Remember to use Documentation/SubmitChecklist when testing your code *** -
Hm, I suppose that table only applies to userspace, not kernel... --- ~Randy *** Remember to use Documentation/SubmitChecklist when testing your code *** -
Doing 64-bit Linux non-LP64 would be an interesting exercise in masochism... -
32-bit userspace?
On 64-bit, `long' is 64-bit on all platforms supported by Linux.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
-
All types are as wide as the compiler makes them. Compiler short int long llong Turbo C 16 16 32 - GCC -m32 16 32 32 64 GCC -m64 16 32 64 64 Jan -- -
We do not support building Linux with Turbo C (or MS Visual C for Win64
P64).
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
-
We're talking about types which are exposed to userspace. -- dwmw2 -
Yes, and all 64-bit software built using kernel headers must be built in LP64 mode, anything else is pure insanity. On LP64 (IE: how the kernel itself is compiled on EVERY 64-bit arch): char == 8 bits short == 16 bits int == 32 bits long == 64 bits pointer == 64 bits long long == 64 bits On LP32 (IE: how the kernel itself is compiled on EVERY 32-bit arch): char == 8 bits short == 16 bits int == 32 bits long == 32 bits pointer == 32 bits long long == 64 bits Ergo we can simply require that if you want to use kernel headers you must be using the same mode as the kernel is compiled in (LP32 or LP64). The simplest guaranteed-not-to-break way to do this on _every_ supported platform is: typedef signed char __s8; typedef unsigned char __s8; typedef signed short __s16; typedef unsigned short __s16; typedef signed int __s32; typedef unsigned int __s32; # if __STDC_VERSION__ >= 19901L typedef signed long long __s64; typedef unsigned long long __s64; # elif defined(__GNUC__) __extension__ typedef signed long long __s64; __extension__ typedef unsigned long long __s64; # endif If you have some other compiler that works under linux *AND* supports a 64-bit type in non-C99-mode (whether "long long" or something else), then they are welcome to submit patches to fix it. Cheers, Kyle Moffett -
Changing this will give you a zillion warnings for printk formats. -Andi -
Why? If this were a problem then we'd be getting a zillion warnings *now* from all the 32-bit archs (which already use "long long" for 64- bit. This would actually make it _easier_ to get the printk formats right, as you could always use %ull for 64-bit types without having to cast for 64-bit platforms. This is another way to get around the "build 32-bit-compat userspace on 64-bit kernel headers" problem: It tells GCC to use the "smallest" available type of the given size, which ends up being exactly the types we use now. On the other hand, it only works for GCC which sort of ruins most of the reason for changing the types in the first place. typedef signed __s8 __attribute__((__mode__(__QI__))); typedef unsigned __u8 __attribute__((__mode__(__QI__))); typedef signed __s16 __attribute__((__mode__(__HI__))); typedef unsigned __u16 __attribute__((__mode__(__HI__))); typedef signed __s32 __attribute__((__mode__(__SI__))); typedef unsigned __u32 __attribute__((__mode__(__SI__))); typedef signed __s64 __attribute__((__mode__(__DI__))); typedef unsigned __u64 __attribute__((__mode__(__DI__))); Cheers, Kyle Moffett -
include/asm-parisc/types.h:typedef unsigned long long __u64; For both 32 and 64-bit. include/asm-sh64/types.h:typedef unsigned long long __u64; include/asm-x86_64/types.h:typedef unsigned long long __u64; So that's three architectures that violate your first assertion. -
Oh, ok, that makes it even easier to say this with certainty: Changing the other 64-bit archs to use "long long" for their 64-bit numbers will not cause additional warnings. I'm also almost certain there are no architectures which use "long long" for 128-bit integers. (Moreover, I can't find hardly anything which does 128-bit integers at all). Cheers, Kyle Moffett -
unsigned long and unsigned long long have the same size, precision and alignment on all LP64 arches, that's true. But they have different ranks and more importantly they mangle differently in C++. So, whether some user exposed type uses unsigned long or unsigned long long is part of the ABI, whether that's size_t, uintptr_t, uint64_t, u_int64_t or any other type, you can't change it without breaking the ABI. Jakub -
That sounds *extraordinarily* broken. Hopefully this would *not* affect the type of a function which is passed a C "struct" containing the "long long", right? Hmm, I guess the question is: Do we support people directly passing __u64 to C++ functions in userspace? I could understand, perhaps, passing around structures defined in the kernel headers, but certainly not the kernel-internal types. The only reason we even export those is so we can have a private set of bit-size-defined types with which to define kernel ABI structures. Cheers, Kyle Moffett -
No, you really don't want to do that, because then u64 != uint64_t on those platforms. -hpa -
the intel compiler does indeed claim to be gcc. rday -- ======================================================================== Robert P. J. Day Linux Consulting, Training and Annoying Kernel Pedantry Waterloo, Ontario, CANADA http://fsdev.net/wiki/index.php?title=Main_Page ======================================================================== -
__extension__ should be macroized if so. However, since something that doesn't have "long long" won't be able to compile those headers, there is absolutely no reason to not simply use it -- you get an error message on that line, which is good enough. There is definitely no reason to use a test like the one above, which is wrong for several supportable compilers. -hpa -
It definitely does hurt when using those types in files that may want to be used by the C library (as opposed to the end user.) However, there is no reason why there should be anything funny about the declaration of those types. -hpa -
???? This uses double negation and is close to unreadable. A kernel include file that defines an interface to a user space program should be self containing (that means that all includes for all non-standard types should be done inside these include files). Whether or not C-99 types are used or not is less important than to use type definitions written I recommend you to install Sun Studio and to try to compile star or cdrtools using Sun Studio by calling "make CCOM_suncc". ftp://ftp.berlios.de/pub/star/alpha/ ftp://ftp.berlios.de/pub/cdrecord/alpha/ You may need to hand edit the file incs/<arch-dir>/{xconfig.h!rules.conf} in order to enable the auto-disabled features. In any case, self reading the error messages from Sun Studio helps more than trying to discuss it. Jörg -- EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin js@cs.tu-berlin.de (uni) schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily -
Yes. In general we try to achieve this. Most header files should include <asm/types.h>, which tends to define the types we use in terms which I have no interest in doing this for myself, and I suspect that if I tried it I'd find that Sun Studio doesn't exist for Linux/PowerPC anyway. Please just show the error messages. -- dwmw2 -
Apart from the usual whining about GNU make, the error message is: make: *** No rule to make target `CCOM_suncc'. Stop. If I actually install smake, as Jörg recommends, the message becomes: smake: Can't find any source for 'CCOM_suncc'. smake: Couldn't make 'CCOM_suncc'. -- Hilsen Harald. -
Well, I was in hope that a small typo (in special as the correct spelling is in
the file README.compile) should not be a problem.
You need to use CCOM=suncc
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
No, I need someone else to use CCOM=suncc for me. Unless suncc works on Linux/PowerPC _and_ you find me a few extra hours in the day to deal with bugs for someone who can't even be bothered to write a coherent bug report. -- dwmw2 -
Joerg.Schilling@fokus.fraunhofer.de (Joerg Schilling) writes: >> If I actually install smake, as J
Mmmmm, if you call "cdrecord -scanbus", what do you get?
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
I may have misunderstood your make system. I cd-ed into the cdrtools directory, ran ./Gmake.linux clean (I had already compiled with GCC), and then ./Gmake.linux CCOM=suncc. I didn't see any errors at the end of the make output. However, what confused me was that the cdrecord binary wasn't removed. And scrolling back, I see several compile errors from Sun's C compiler. Shouldn't "make clean" remove the executable files? -- Hilsen Harald. -
make clean only removes the *.o files.
If you call "make relink", all results are removed before recreating.
BTW: your problem is caused by a "bash" bug. With a shell that correctly
works with "sh -ce 'command...'", you would see an abort after the first
error.
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
Well, there is no need to go to ANSI C as pre-ANSI would to it also.
Using plain C in the Linux kernel include files would be sufficient for all
compilers that make sense to be supported.
Jörg
--
EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
js@cs.tu-berlin.de (uni)
schilling@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
-
