Re: [PATCH] build: avoid use of C11 specific %t format size spec
On 02/08/2023 00:11, Paul Eggert wrote:
On 2023-08-01 06:28, Pádraig Brady wrote:
perhaps we can now use %j, %t, and %z now?
I hope so, particularly as we haven't seen any issues with Coreutils 9.1
or later.
I haven't run into problems using these printf formats for a while.
Gnulib doc is ambiguous, but suggests it could be an issue on AIX 5.1,
HP-UX 11.23, IRIX 6.5, Solaris 9, Cygwin 1.5.24, mingw, MSVC 14. Of
these, the only platforms that are not obsolete are mingw and MSVC 14,
and Stack Overflow[1] suggests that this is only with older MS-Windows
versions and can be worked around easily on those older versions by
whoever's doing the port.
Diffutils doesn't work on MS-Windows now anyway, as per Bug#64811, so
I'll cc to that bug report to give Gisle Vanem a heads-up about this
other issue.
[1]:
https://stackoverflow.com/questions/44382862/how-to-printf-a-size-t-without-warning-in-mingw-w64-gcc-7-1
Thanks for the info.
I'll apply the following to allow use of all C99 size specs so.
cheers,
PádraigFrom 322530bd433996f8a4f2086c0229facf41c6c709 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C3=A1draig=20Brady?=
Date: Sat, 12 Aug 2023 11:21:50 +0100
Subject: [PATCH] maint: allow use of printf C99 integer size specifiers
Older systems that had issues with these like HPUX and Solaris 8
are no longer supported, or can apply patches to provide support.
Also we've used %td since coreutils 9.1, with no reported issues.
* cfg.mk (sc_prohibit-c99-printf-format): Remove to allow use of
%[jtz] size specifiers, which allows for cleaner code
by avoiding the need to cast to PRI?MAX etc.
---
cfg.mk | 11 ---
1 file changed, 11 deletions(-)
diff --git a/cfg.mk b/cfg.mk
index 630251b65..6db0daf71 100644
--- a/cfg.mk
+++ b/cfg.mk
@@ -299,17 +299,6 @@ sc_check-I18N-AUTHORS:
'$$(LIBICONV)' 1>&2; exit 1; }; \
done
-# Disallow the C99 printf size specifiers %z and %j as they're not portable.
-# The gnulib printf replacement does support them, however the printf
-# replacement is not currently explicitly depended on by the gnulib error()
-# module for example. Also we use fprintf() in a few places to output simple
-# formats but don't use the gnulib module as it is seen as overkill at present.
-# We'd have to adjust the above gnulib items before disabling this.
-sc_prohibit-c99-printf-format:
- @cd $(srcdir)/src && GIT_PAGER= git grep -n '%[0*]*[jz][udx]' *.c\
- && { echo '$(ME): Use PRI*MAX instead of %j or %z' 1>&2; exit 1; } \
- || :
-
# Ensure the alternative __attribute (keyword) form isn't used as
# that form is not elided where required. Also ensure that we don't
# directly use attributes already defined by gnulib.
--
2.41.0
Re: [PATCH] build: avoid use of C11 specific %t format size spec
On 2023-08-01 06:28, Pádraig Brady wrote: perhaps we can now use %j, %t, and %z now? I hope so, particularly as we haven't seen any issues with Coreutils 9.1 or later. I haven't run into problems using these printf formats for a while. Gnulib doc is ambiguous, but suggests it could be an issue on AIX 5.1, HP-UX 11.23, IRIX 6.5, Solaris 9, Cygwin 1.5.24, mingw, MSVC 14. Of these, the only platforms that are not obsolete are mingw and MSVC 14, and Stack Overflow[1] suggests that this is only with older MS-Windows versions and can be worked around easily on those older versions by whoever's doing the port. Diffutils doesn't work on MS-Windows now anyway, as per Bug#64811, so I'll cc to that bug report to give Gisle Vanem a heads-up about this other issue. [1]: https://stackoverflow.com/questions/44382862/how-to-printf-a-size-t-without-warning-in-mingw-w64-gcc-7-1
Re: [PATCH] build: avoid use of C11 specific %t format size spec
%t is C99 -- Andreas Schwab, sch...@linux-m68k.org GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1 "And now for something completely different."
Re: [PATCH] build: avoid use of C11 specific %t format size spec
On 01/08/2023 14:07, Pádraig Brady wrote: * cfg.mk (sc_prohibit-c11-printf-format): Renamed and updated from sc_prohibit-c99-printf-format. * src/dd.c: Adjust %td -> %"PRIdMAX". * src/dircolors.c: Likewise. * src/du.c: Likewise. * src/join.c: Likewise. * src/od.c: Likewise. * src/pathchk.c: Likewise. Actually I see that we've used %td since coreutils 9.1 I'm surprised we've not hit build issues with that, but perhaps we can now use %j, %t, and %z now? cheers, Pádraig
[PATCH] build: avoid use of C11 specific %t format size spec
* cfg.mk (sc_prohibit-c11-printf-format): Renamed and updated
from sc_prohibit-c99-printf-format.
* src/dd.c: Adjust %td -> %"PRIdMAX".
* src/dircolors.c: Likewise.
* src/du.c: Likewise.
* src/join.c: Likewise.
* src/od.c: Likewise.
* src/pathchk.c: Likewise.
---
cfg.mk | 8
src/dd.c| 17 +
src/dircolors.c | 8
src/du.c| 10 ++
src/join.c | 3 ++-
src/od.c| 4 ++--
src/pathchk.c | 9 +
7 files changed, 32 insertions(+), 27 deletions(-)
diff --git a/cfg.mk b/cfg.mk
index 630251b65..1a0affd4b 100644
--- a/cfg.mk
+++ b/cfg.mk
@@ -299,15 +299,15 @@ sc_check-I18N-AUTHORS:
'$$(LIBICONV)' 1>&2; exit 1; }; \
done
-# Disallow the C99 printf size specifiers %z and %j as they're not portable.
+# Disallow the C11 printf size specifiers %z, %j and %t as they're not
portable.
# The gnulib printf replacement does support them, however the printf
# replacement is not currently explicitly depended on by the gnulib error()
# module for example. Also we use fprintf() in a few places to output simple
# formats but don't use the gnulib module as it is seen as overkill at present.
# We'd have to adjust the above gnulib items before disabling this.
-sc_prohibit-c99-printf-format:
- @cd $(srcdir)/src && GIT_PAGER= git grep -n '%[0*]*[jz][udx]' *.c\
- && { echo '$(ME): Use PRI*MAX instead of %j or %z' 1>&2; exit 1; } \
+sc_prohibit-c11-printf-format:
+ @cd $(srcdir)/src && GIT_PAGER= git grep -n '%[0*]*[tjz][udx]' *.c \
+ && { echo '$(ME): Use PRI*MAX instead of %j, %t or %z' 1>&2;exit 1; }\
|| :
# Ensure the alternative __attribute (keyword) form isn't used as
diff --git a/src/dd.c b/src/dd.c
index 21f7a4ede..ecf678780 100644
--- a/src/dd.c
+++ b/src/dd.c
@@ -675,8 +675,9 @@ alloc_ibuf (void)
{
char hbuf[LONGEST_HUMAN_READABLE + 1];
error (EXIT_FAILURE, 0,
- _("memory exhausted by input buffer of size %td bytes (%s)"),
- input_blocksize,
+ _("memory exhausted by input buffer of size %"PRIdMAX
+ " bytes (%s)"),
+ (intmax_t) input_blocksize,
human_readable (input_blocksize, hbuf,
human_opts | human_base_1024, 1, 1));
}
@@ -697,9 +698,9 @@ alloc_obuf (void)
{
char hbuf[LONGEST_HUMAN_READABLE + 1];
error (EXIT_FAILURE, 0,
- _("memory exhausted by output buffer of size %td"
+ _("memory exhausted by output buffer of size %"PRIdMAX
" bytes (%s)"),
- output_blocksize,
+ (intmax_t) output_blocksize,
human_readable (output_blocksize, hbuf,
human_opts | human_base_1024, 1, 1));
}
@@ -1136,9 +1137,9 @@ iread (int fd, char *buf, idx_t size)
{
idx_t prev = prev_nread;
if (status_level != STATUS_NONE)
-diagnose (0, ngettext (("warning: partial read (%td byte); "
+diagnose (0, ngettext (("warning: partial read (%"PRIdMAX" byte); "
"suggest iflag=fullblock"),
- ("warning: partial read (%td bytes); "
+ ("warning: partial read (%"PRIdMAX" bytes);
"
"suggest iflag=fullblock"),
select_plural (prev)),
prev);
@@ -2484,8 +2485,8 @@ main (int argc, char **argv)
error (EXIT_FAILURE, 0,
_("offset too large: "
"cannot truncate to a length of seek=%"PRIdMAX""
- " (%td-byte) blocks"),
- seek_records, output_blocksize);
+ " (%"PRIdMAX"-byte) blocks"),
+ seek_records, (intmax_t) output_blocksize);
/* Open the output file with *read* access only if we might
need to read to satisfy a 'seek=' request. If we can't read
diff --git a/src/dircolors.c b/src/dircolors.c
index 8a86efb76..a47ff95e8 100644
--- a/src/dircolors.c
+++ b/src/dircolors.c
@@ -322,8 +322,8 @@ dc_parse_stream (FILE *fp, char const *filename)
if (arg == nullptr)
{
- error (0, 0, _("%s:%td: invalid line; missing second token"),
- quotef (filename), line_number);
+ error (0, 0, _("%s:%"PRIdMAX": invalid line; missing second token"),
+ quotef (filename), (intmax_t) line_number);
ok = false;
free (keywd);
continue;
@@ -377,9 +377,9 @@ dc_parse_stream (FILE *fp, char const *filename)
if (unrecognized && (state == ST_TERMSURE || state == ST_TERMYES))
{
- error (0, 0, _("%s:%td: unrecognized keyword %s"),
+ error (0, 0, _("%s:%"PRIdMAX": unrecognized keyword %s"),
(filename ?
