Re: [PATCH 21/21] bootstrap: fix 'find: missing argument to `-exec''

Bruno Haible wrote: So what you are doing here is to add a stray ';', which leads to a syntax error: I guess the patch is attempting to work around an incompatibility with the 'find' implementation on OS/2, which evidently does not support the '+' syntax that was introduced in POSIX:2001. Bu

Re: [PATCH 21/21] bootstrap: fix 'find: missing argument to `-exec''

> -&& ! find "$local_gl_dir" -name '*.diff' -exec false {} +; then > +&& ! find "$local_gl_dir" -name '*.diff' -exec false {} + ';'; then Sorry, this does not look right. According to POSIX [1] the '+' argument terminates the primary expression "-exec false {} +". So what you are doing her

Re: [PATCH 15/21] freopen: workaround freopen() on OS/2 kLIBC

> +#ifdef __KLIBC__ > + FILE *result = freopen (filename, mode, stream); > + > + /* On OS/2 kLIBC, freopen() returns NULL even if it is successful > + if filename is NULL. */ > + if (!result && !errno) > +result = stream; 1) In the comment you say 'if filename is NULL' but in the code y

Re: [PATCH 12/21] pipe_filter_ii_execute: port to OS/2 kLIBC

> * lib/pipe-filter-ii.c (start_wrapper, _beginthreadex, CloseHandle, > WaiForSingleObject, WaitForMultipleObjects): New on OS/2 kLIBC. Since you make use of these functions in w32spawn.h, their implementation does not belong in pipe-filter-ii.c, but rather in some other .c file, say w32-on-os2.c.

Re: [PATCH 13/21] w32spawn: clear SHELL_SPECIAL_CHARS and SHELL_SPACE_CHAR on OS/2 kLIBC

> +#ifndef __KLIBC__ > #define SHELL_SPECIAL_CHARS "\"\\ > \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020\021\022\023\024\025\026\027\030\031\032\033\034\035\036\037*?" > #define SHELL_SPACE_CHARS " > \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020\021\022\023

Re: [PATCH 10/21] binary-io: put fd in binary mode if it is not a console on EMX

> * lib/binary-io.h (SET_BINARY): put fd in binary mode if it is not a > console on EMX. A changelog entry should usually specify the difference between the previous behaviour and the new behaviour. Therefore here it should better read: * lib/binary-io.h (SET_BINARY): Don't put fd in binary mode

Re: [PATCH 07/21] find_executable: port to EMX

> + DosGetInfoBlocks (NULL, &ppib); Reliability: Please don't ignore the return value of DosGetInfoBlocks. Also, if possible, can you please add a comment with the URL of the specification of DosGetInfoBlocks? Future readers of this code should be able to understand its parameters. Bruno

Re: [PATCH 08/21] get_shared_library_fullname: port to EMX

> +#ifdef __EMX__ > +# define INCL_DOS > +# include > + > +#define strcmp stricmp > +#define strncmp strnicmp > +#endif Please indent these lines: # define strcmp stricmp # define strncmp strnicmp > +if (DosQueryModuleName (hModule, sizeof (location), location)) Can you add a comment

Re: [PATCH 05/21] spawn: do not include sched.h on OS/2 kLIBC

> * lib/spawn.in.h: Do not include sched.h on OS/2 kLIBC. POSIX [1] specifies that must declare the tag 'sched_param'. Where do you want to get it from, if not from ? Bruno [1] http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/spawn.h.html

Re: [PATCH 01/21] gnulib-tool: disable a symbolic link on OS/2

> (do_copyrights): Clear unless no_symlink_support is set. >-S | --more-symlinks | --more-symlink | --more-symlin | --more-symli | > --more-syml | --more-sym | --more-sy | --more-s | --more- | --more | --mor | > --mo ) > -symbolic=true > -do_copyrights= > +test -z

Re: [PATCH 20/21] fdopendir: port to OS/2 kLIBC

KO Myung-Hun wrote: * lib/fdopendir.c (fdopendir): Implement on OS/2 kLIBC. Patches from coreutils 8.8 by Paul Smedley. First, we don't have copyright assignment from Paul Smedley, so we can't install the patch without having that cleared up one way or another. Second, the patch seems to in

Re: [PATCH 19/21] utimens: define HAVE_WORKING_UTIMES to 1 on OS/2 kLIBC

KO Myung-Hun wrote: +#ifdef __KLIBC__ /* fs doesn't to better than seconds, but we must use futimes! */ +# define HAVE_WORKING_UTIMES 1 +#endif Sorry, I don't follow. Is the problem that kLIBC utimes operates on file systems whose time stamps resolution are 2 seconds? I recall that this is

Re: [PATCH 18/21] Workaround for dirfd() on OS/2 kLIBC

KO Myung-Hun wrote: * lib/fts.c (fts_build): Do not check dir_fd on OS/2 kLIBC. I'm afraid this change needs more explanation. If the problem is that dirfd returns -2, the earlier part of fts_build will return when that happens, so that problem cannot be fixed by the patch that you sent. W

Re: [PATCH 17/21] Workaround for D_INO_IN_DIRENT on OS/2 kLIBC

KO Myung-Hun wrote: d_ino is not supported correctly on OS/2 kLIBC 0.6. This will be fixed in 0.7. How about testing for the bug in 'configure', then? That way, D_INO_IN_DIRENT will be correct. Programs other than Gnulib use it, so that'd be a better approach.

Re: [PATCH 16/21] wcwidth: fix 'conflicting types' error for `__wcwidth' on OS/2 kLIBC

KO Myung-Hun wrote: +#ifdef __KLIBC__ +/* To avoid 'conflicting types' error for `__wcwidth' on OS/2 kLIBC. + wchar_t(unsigned short) is defined differently from wint_t(int) on + OS/2 kLIBC. */ +# undef wcwidth +/* Ignore wcwidth() of OS/2 kLIBC */ +# undef HAVE_WCWIDTH +#endif The #undef i

Re: [PATCH 15/21] freopen: workaround freopen() on OS/2 kLIBC

I don't see why the workaround must be in freopen-safer.c. freopen-safer has freopen as a prerequisite, so it can assume that freopen doesn't have the bug in question. +#ifdef __KLIBC__ + FILE *result = freopen (filename, mode, stream); + + /* On OS/2 kLIBC, freopen() returns NULL even if i

Re: [PATCH 14/21] getdtablesize: do not use getrlimit() on OS/2 kLIBC

KO Myung-Hun wrote: getrlimit() was implemented in kLIBC and works well. Strangely, however, it crashes due to 'stack overflow' or 'SIGSEGV' when used in GNU M4. This needs explanation in the form of a brief comment in the source code. Also, I don't see how it can be said to work "well" if it

Re: [PATCH 13/21] w32spawn: clear SHELL_SPECIAL_CHARS and SHELL_SPACE_CHAR on OS/2 kLIBC

KO Myung-Hun wrote: +#ifndef __KLIBC__ #define SHELL_SPECIAL_CHARS "\"\\ \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020\021\022\023\024\025\026\027\030\031\032\033\034\035\036\037*?" #define SHELL_SPACE_CHARS " \001\002\003\004\005\006\007\010\011\012\013\014\015\016\017\020

Re: [PATCH 12/21] pipe_filter_ii_execute: port to OS/2 kLIBC

KO Myung-Hun wrote: +#if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__ || defined __KLIBC__ This patch also has some lines that are too long and should be broken up.

Re: [PATCH 08/21] get_shared_library_fullname: port to EMX

KO Myung-Hun wrote: +#if !((defined _WIN32 || defined __WIN32__) && !defined __CYGWIN__) && !defined __EMX__ This line is longer than 80 characters ; please break it up, e.g., like this: #if (!((defined _WIN32 || defined __WIN32__) && !defined __CYGWIN__) \ && !defined __EMX__) I know s

Re: [PATCH 05/21] spawn: do not include sched.h on OS/2 kLIBC

KO Myung-Hun wrote: /* Get definitions of 'struct sched_param' and 'sigset_t'. But avoid namespace pollution on glibc systems. */ #if !(defined __GLIBC__ && !defined __UCLIBC__) -# include +# ifndef __KLIBC__ +# include +# endif Please add a brief note to that comment, to explain

Re: [PATCH 04/21] stdint: typedef gl_intptr_t and gl_uintptr_t correctly on OS/2 kLIBC

KO Myung-Hun wrote: +#ifndef __KLIBC__ typedef long int gl_intptr_t; typedef unsigned long int gl_uintptr_t; +#else +typedef int gl_intptr_t; +typedef unsigned int gl_uintptr_t; +#endif This looks brittle. First, why is it used at all? That is, what's wrong with kLIBC's stdint.h, which c

Re: [PATCH 03/21] git-version-gen: remove CR as well

KO Myung-Hun wrote: +echo "$v" | tr -d "$cr$nl" How about this instead? printf '%s' "$v" This avoids the need for cr and for tr, which is simpler and a bit faster. The script is already using printf so this shouldn't be a portability issue.

Re: [PATCH 02/21] gnulib-tool: recognize x:* as an absolute path

KO Myung-Hun wrote: * gnulib-tool (func_gnulib_dir): Add [A-Za-z]:* case. The usual globbing pattern elsewhere in Gnulib is '?:*', not '[A-Za-z]:*'. This is simpler and is just as good in practice.

Re: [PATCH 01/21] gnulib-tool: disable a symbolic link on OS/2

KO Myung-Hun wrote: +# OS/2 does not support a symbolc link at all +case `uname -s` in + "OS/2") no_symlink_support=true;; + *) no_symlink_support=;; +esac Surely it would be better to test this directly, by running 'ln -s' and seeing it fail, than to look at operating system names. 'symbo

[PATCH] do not try to translate empty pre doc in argp_help

Every po file seems to include entry with empty msgid with some meta information about it. If argp->doc string contains empty pre section (like "\v" "Some text to output after options"), argp_help tries to translate empty string and emits this meta information, like in bor@opensuse:~/src/grub> gru

Re: gnulib - test-getcwd failure on Mac OS X

For context, gnulib-tests/test-getcwd.c is failing on Mac OS X with exit(7)... On 30/11/14 05:59, Assaf Gordon wrote: > One more thing regarding AT_FDCWD: > > On Nov 30, 2014, at 0:54, Assaf Gordon wrote: > >> >> But I did notice this: >> In “m4/getcwd-path-max.m4” there is C code which is very

[PATCH 01/21] gnulib-tool: disable a symbolic link on OS/2

OS/2 does not support a symbolic link at all. * gnulib-tool (no_symlink_support): New. Set to true on OS/2, otherwise clear. (symbolic, lsymbolic): Set to true unless no_symlink_support is set. (do_copyrights): Clear unless no_symlink_support is set. --- gnulib-tool | 14 ++ 1 file ch

[PATCH 19/21] utimens: define HAVE_WORKING_UTIMES to 1 on OS/2 kLIBC

configure test for HAVE_WORKING_UTIMES fails on OS/2 kLIBC. But it is required. * lib/utimesns.c (HAVE_WORKING_UTIMES): Define to 1 on OS/2 kLIBC. Patches from coreutils 8.8 by Paul Smedley. --- lib/utimens.c | 4 1 file changed, 4 insertions(+) diff --git a/lib/utimens.c b/lib/utimens.c i

[PATCH 21/21] bootstrap: fix 'find: missing argument to `-exec''

* build-aux/bootstrap: Append quoted ; to '-exec' command. --- build-aux/bootstrap | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/build-aux/bootstrap b/build-aux/bootstrap index 4f0493a..761d25a 100755 --- a/build-aux/bootstrap +++ b/build-aux/bootstrap @@ -544,7 +544,7 @@ esa

[PATCH 02/21] gnulib-tool: recognize x:* as an absolute path

On OS/2, x:* is an absolute path, too. * gnulib-tool (func_gnulib_dir): Add [A-Za-z]:* case. (func_relconcat): Likewise. --- gnulib-tool | 8 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/gnulib-tool b/gnulib-tool index f1c91c6..b0b77e3 100755 --- a/gnulib-tool +++ b/gnul

[PATCH 15/21] freopen: workaround freopen() on OS/2 kLIBC

On OS/2 kLIBC, freopen() returns NULL even if it is successful if filename is NULL. * lib/freopen-safer.c (klibc_freopen): New. * lib/freopen.c (orig_freopen): Workaround. --- lib/freopen-safer.c | 19 +++ lib/freopen.c | 11 +++ 2 files changed, 30 insertions(+) di

[PATCH 20/21] fdopendir: port to OS/2 kLIBC

* lib/fdopendir.c (fdopendir): Implement on OS/2 kLIBC. Patches from coreutils 8.8 by Paul Smedley. --- lib/fdopendir.c | 57 + 1 file changed, 57 insertions(+) diff --git a/lib/fdopendir.c b/lib/fdopendir.c index b6c94a0..efe9c7a 100644 --

[PATCH 17/21] Workaround for D_INO_IN_DIRENT on OS/2 kLIBC

d_ino is not supported correctly on OS/2 kLIBC 0.6. This will be fixed in 0.7. * lib/backupfile.c (REAL_DIR_ENTRY): Define to 1 on OS/2 kLIBC. * lib/fts.c (D_INO): Define to NOT_AN_INODE_NUMBER on OS/2 kLIBC. * lib/getcwd.c (MATCHING_INO): Define to true on OS/2 kLIBC. Patches from coreutils 8.8

[PATCH 08/21] get_shared_library_fullname: port to EMX

* lib/relocatable.c: Define strcmp and strncmp to stricmp and strnicmp on EMX, respectively. (_DLL_InitTerm): New on EMX. (get_shared_library_fullname): Implement on EMX. --- lib/relocatable.c | 51 --- 1 file changed, 48 insertions(+), 3 deletions(-

[PATCH] OS/2 patches

Hi/2. These are OS/2 patches. Review, please... [PATCH 01/21] gnulib-tool: disable a symbolic link on OS/2 [PATCH 02/21] gnulib-tool: recognize x:* as an absolute path [PATCH 03/21] git-version-gen: remove CR as well [PATCH 04/21] stdint: typedef gl_intptr_t and gl_uintptr_t correctly [PATCH 05/

[PATCH 18/21] Workaround for dirfd() on OS/2 kLIBC

dirfd() always return -2 on OS/2 kLIBC 0.6. This will be fixed in 0.7. * lib/fts.c (fts_build): Do not check dir_fd on OS/2 kLIBC. Patches from coreutils 8.8 by Paul Smedley. --- lib/fts.c | 4 1 file changed, 4 insertions(+) diff --git a/lib/fts.c b/lib/fts.c index 7f3cdc7..02ebdde 100644

[PATCH 07/21] find_executable: port to EMX

* lib/progreloc.c (find_executable): Implement on EMX. --- lib/progreloc.c | 18 ++ 1 file changed, 18 insertions(+) diff --git a/lib/progreloc.c b/lib/progreloc.c index 3d7b6a9..5b9e247 100644 --- a/lib/progreloc.c +++ b/lib/progreloc.c @@ -44,6 +44,12 @@ # include #endif +#

[PATCH 13/21] w32spawn: clear SHELL_SPECIAL_CHARS and SHELL_SPACE_CHAR on OS/2 kLIBC

spawn() on OS/2 kLIBC is not silly like one on Windows * libc/w32spawn.h (SHELL_SPECIAL_CHARS, SHELL_SPACE_CHAR): Set both to empty string on OS/2 kLIBC. --- lib/w32spawn.h | 5 + 1 file changed, 5 insertions(+) diff --git a/lib/w32spawn.h b/lib/w32spawn.h index 5b7f7ef..3a37092 100644 --- a

[PATCH 09/21] relocatable: support UNIXROOT in relocate() on EMX

UNIXROOT is used to specify a drive of a root of FHS. So if a path is started with '/', then it should be translated to "$UNIXROOT/". * lib/relocatable.c (relocate): Prepend $UNIXROOT to pathname if it is started with '/' on EMX. --- lib/relocatable.c | 21 + 1 file changed, 2

[PATCH 05/21] spawn: do not include sched.h on OS/2 kLIBC

kLIBC declares struct sched_param in spawn.h, so a redefinition error occurs. * lib/spawn.in.h: Do not include sched.h on OS/2 kLIBC. --- lib/spawn.in.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/lib/spawn.in.h b/lib/spawn.in.h index eb76bb0..134cb08 100644 --- a/lib/s

[PATCH 04/21] stdint: typedef gl_intptr_t and gl_uintptr_t correctly on OS/2 kLIBC

intptr_t and uintptr_t are defined as int and unsigned int on OS/2 kLIBC, respectively. This fixes a type conflict of sbrk() on OS/2 kLIBC. * lib/stdint.in.h (gl_intptr_t, gl_uintptr_t): typedef to int and unsigned int on OS/2 kLIBC, repsectively. --- lib/stdint.in.h | 5 + 1 file changed, 5

[PATCH 14/21] getdtablesize: do not use getrlimit() on OS/2 kLIBC

getrlimit() was implemented in kLIBC and works well. Strangely, however, it crashes due to 'stack overflow' or 'SIGSEGV' when used in GNU M4. So just use getdtablesize(). * lib/getdtablesize.c (rpl_getdtablesize): Do not use getrlimit() on OS/2 kLIBC. --- lib/getdtablesize.c | 2 ++ 1 file change

[PATCH 16/21] wcwidth: fix 'conflicting types' error for `__wcwidth' on OS/2 kLIBC

wchar_t(unsigned short) is defined differently from wint_t(int) on OS/2 kLIBC. * lib/wcwidth.c (wcwidth, HAVE_WCWIDTH): Undefine on OS/2 kLIBC. --- lib/wcwidth.c | 9 + 1 file changed, 9 insertions(+) diff --git a/lib/wcwidth.c b/lib/wcwidth.c index d7837bb..b36abd5 100644 --- a/lib/wcwi

[PATCH 03/21] git-version-gen: remove CR as well

On OS/2, a new line consists of CR and LF. * build-aux/git-version-gen: Remove CR as well. --- build-aux/git-version-gen | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/build-aux/git-version-gen b/build-aux/git-version-gen index 47d6576..77fb2ad 100755 --- a/build-aux/git-ve

[PATCH 12/21] pipe_filter_ii_execute: port to OS/2 kLIBC

Pipes on kLIBC does not support O_NONBLOCK like Win32. * lib/pipe-filter-ii.c (start_wrapper, _beginthreadex, CloseHandle, WaiForSingleObject, WaitForMultipleObjects): New on OS/2 kLIBC. Reuse Win32 codes on OS/2 kLIBC. * lib/spawn-pipe.c: Reuse Win32 codes on OS/2 kLIBC. * lib/w32spawn.h: Do not

[PATCH 06/21] Fix character encoding aliases for OS/2

* lib/config.charset: Remove os2* from case "$os" in * lib/localcharset.c (get_charset_aliases): Use embedded encoding aliases on OS/2. --- lib/config.charset | 4 +--- lib/localcharset.c | 60 +- 2 files changed, 60 insertions(+), 4 deletions(-

[PATCH 10/21] binary-io: put fd in binary mode if it is not a console on EMX

* lib/binary-io.h (SET_BINARY): put fd in binary mode if it is not a console on EMX. --- lib/binary-io.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/lib/binary-io.h b/lib/binary-io.h index 7928f8c..696cdf9 100644 --- a/lib/binary-io.h +++ b/lib/binary-io.h @@ -60,7 +60,7 @@

[PATCH 11/21] pipe-filter-aux: undefine HAVE_SELECT on KLIBC

On OS/2 kLIBC, select() works only on sockets. * lib/pipe-filter-aux.h (HAVE_SELECT): Undefine on OS/2 kLIBC. --- lib/pipe-filter-aux.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/lib/pipe-filter-aux.h b/lib/pipe-filter-aux.h index 8f2a707..ee63ac8 100644 --- a/lib/pi