some module inconsistencies

2007-01-28 Thread Ralf Wildenhues
Hello Bruno, all, gnulib-tool --with-tests --test currently gives me | gnulib-tool: module utf8-ucs4-safe doesn't exist | gnulib-tool: module utf16-ucs4-safe doesn't exist I suppose the renaming session isn't over yet? ;-) (Wow, you're productive these days!) Cheers, Ralf

small gnulib-tool nits

2007-01-28 Thread Ralf Wildenhues
Some more bits in gnulib-tool depend upon the locale. The patch below fixes some of those. It also kills one superfluous grep, and avoids a misparse of the ACLOCAL_AMFLAGS. OK to apply? Cheers, Ralf 2007-01-28 Ralf Wildenhues [EMAIL PROTECTED] * gnulib-tool

gnulib-tool and line length: gl_LIBOBJS

2007-01-28 Thread Ralf Wildenhues
The combined length of object names in gl_LIBOBJS can get pretty close to the line limit of some 'make' implementations, or, FWIW, the portable text file format limit. As Autoconf 2.60 allows newlines in substituted values, there's an easy fix for users of new Autoconf. WDYT? Cheers, Ralf

Re: some module inconsistencies

2007-01-28 Thread Bruno Haible
Hello Ralf, gnulib-tool --with-tests --test currently gives me | gnulib-tool: module utf8-ucs4-safe doesn't exist | gnulib-tool: module utf16-ucs4-safe doesn't exist Fixed now. Thanks. I suppose the renaming session isn't over yet? ;-) It was supposed to be over. ;-) I corrected a

Re: small gnulib-tool nits

2007-01-28 Thread Bruno Haible
Ralf Wildenhues wrote: 2007-01-28 Ralf Wildenhues [EMAIL PROTECTED] * gnulib-tool (func_emit_lib_Makefile_am, func_add_or_update) (func_create_testdir): Ensure C locale for `grep' and `tr' character ranges. (func_create_megatestdir): Avoid one `grep'. Fix bug in

Re: gnulib-tool and line length: gl_LIBOBJS

2007-01-28 Thread Bruno Haible
Ralf Wildenhues wrote: The combined length of object names in gl_LIBOBJS can get pretty close to the line limit of some 'make' implementations, How small is this limit? or, FWIW, the portable text file format limit. What is this limit? I can edit text files with 1 characters per line in

Re: small gnulib-tool nits

2007-01-28 Thread Ralf Wildenhues
* Bruno Haible wrote on Sun, Jan 28, 2007 at 02:40:37PM CET: Ralf Wildenhues wrote: 2007-01-28 Ralf Wildenhues [EMAIL PROTECTED] * gnulib-tool (func_emit_lib_Makefile_am, func_add_or_update) (func_create_testdir): Ensure C locale for `grep' and `tr' character ranges.

Re: gnulib-tool and line length: gl_LIBOBJS

2007-01-28 Thread Bruno Haible
Ralf Wildenhues wrote: How small is this limit? Around 20kB: http://www.gnu.org/software/autoconf/manual/html_node/Long-Lines-in-Makefiles.html OK, and since gnulib's LIBOBJs line is currently at most 2.4 KB, we can ignore the problem for quite some time. $ grep -h AC_LIBOBJ m4/*.m4 | sed

simplification of unistd.h creation

2007-01-28 Thread Bruno Haible
Hi, The AC_REQUIRE technique just developed for string.h can also be used to simplify the autoconf macros for creating the unistd.h replacement. I committed this. 2007-01-28 Bruno Haible [EMAIL PROTECTED] * m4/unistd_h.m4 (gl_HEADER_UNISTD_DEFAULTS): New macro.

avoid creating sys/time.h replacement when possible

2007-01-28 Thread Bruno Haible
Hi Paul, Jim, Most developers are using glibc systems nowadays. When gnulib creates many replacement header files on such systems, it makes look gnulib a bit ridiculous. So can we avoid to create a header file when it's easy to do so? Here is a proposed patch. 2007-01-28 Bruno Haible [EMAIL

Re: avoid creating sys/time.h replacement when possible

2007-01-28 Thread Jim Meyering
Bruno Haible [EMAIL PROTECTED] wrote: Jim Meyering wrote: - use $(SYS_TIME_H) also in the definition of MOSTLYCLEANFILES This is not a good change. Suppose a user does in the same directory - first, a ./configure; make for a deficient platform, - then, a ./configure for a glibc

Re: build failures: string.h vs. HP-UX

2007-01-28 Thread Paul Eggert
[EMAIL PROTECTED] (Bob Proulx) writes: The problem is that regex.h is defining __restrict to be restrict! ... And of course config.h is defining restrict to be __restrict. #define restrict __restrict That combination is obviously not good. Thanks for reporting that. I installed this,

Re: build failures: string.h vs. HP-UX

2007-01-28 Thread Paul Eggert
Jim Meyering [EMAIL PROTECTED] writes: Avoid a compile error from HP-UX's ia64 cc: s/__restrict\/restrict/ * lib/regex.h: Use restrict, not __restrict, since only the former works with 'config.h's #define restrict Problem reported by Bob Proulx. Arg, these

Re: build failures: string.h vs. HP-UX

2007-01-28 Thread Jim Meyering
Paul Eggert [EMAIL PROTECTED] wrote: Jim Meyering [EMAIL PROTECTED] writes: If you're serious about merging back to glibc, it'll help (name space cleanliness) to add the __ prefix: i.e., change your Restrict to __Restrict. But the prefix _R suffices, no? As far as the standard C name space

new module mpsort for faster sorting

2007-01-28 Thread Paul Eggert
For quite some time Djamel Belazzougui has been suggesting speed improvements to glibc's qsort function. I'm not sure yet that this is a good idea, but it does seem to me that there is use for a function that can sort a vector of pointers to data (as opposed to qsort, which sorts an array of