Dominique Pelle wrote:
> I've compiled Vim-7.3e (2539:9397d2d76340) with the clang compiler on Linux.
>
> All tests pass. However, I see something wrong when doing:
>
> $ vim -c 'saveas! /tmp/foo'
>
> Vim prints:
>
> "/tmp/foo" [New File] 0 lines, 585937717901131776 characters written
>
> Obviously the number of characters written is not correct (should be 0).
>
> Valgrind also reports:
>
> ==18757== Conditional jump or move depends on uninitialised value(s)
> ==18757== at 0x496935B: _itoa (_itoa.c:442)
> ==18757== by 0x496CC85: vfprintf (vfprintf.c:1613)
> ==18757== by 0x498B70B: vsprintf (iovsprintf.c:43)
> ==18757== by 0x49741DA: sprintf (sprintf.c:34)
> ==18757== by 0x80D850C: msg_add_lines (fileio.c:5246)
> ==18757== by 0x80DB5EC: buf_write (fileio.c:4885)
> ==18757== by 0x80A24B0: do_write (ex_cmds.c:2706)
> ==18757== by 0x80A256B: ex_write (ex_cmds.c:2519)
> ==18757== by 0x80BA4D8: do_one_cmd (ex_docmd.c:2656)
> ==18757== by 0x80B739F: do_cmdline (ex_docmd.c:1122)
> ==18757== by 0x80B7F7E: do_cmdline_cmd (ex_docmd.c:728)
> ==18757== by 0x8100D64: exe_commands (main.c:2807)
> ==18757== by 0x80FE1FC: main (main.c:885)
> ==18757== Uninitialised value was created by a stack allocation
> ==18757== at 0x80D8494: msg_add_lines (fileio.c:5238)
>
> And while compiling, I see this warning which points to the same problem:
>
> fileio.c:5248:13: warning: conversion specifies type 'long long' but
> the argument has type 'off_t' (aka 'long')
> [-Wformat]
> "%ldL, %lldC", lnum, nchars
> ~~~^~ ~~~~~~
>
> The number of characters is printed in fileio.c:5248:
>
> 5233 void
> 5234 msg_add_lines(insert_space, lnum, nchars)
> 5235 int insert_space;
> 5236 long lnum;
> 5237 off_t nchars;
> 5238 {
> 5239 char_u *p;
> 5240
> 5241 p = IObuff + STRLEN(IObuff);
> 5242
> 5243 if (insert_space)
> 5244 *p++ = ' ';
> 5245 if (shortmess(SHM_LINES))
> 5246 sprintf((char *)p,
> 5247 #ifdef LONG_LONG_OFF_T
> !5248 "%ldL, %lldC", lnum, nchars
> 5249 #else
> 5250 /* Explicit typecast avoids warning on Mac OS X 10.6 */
> 5251 "%ldL, %ldC", lnum, (long)nchars
> 5252 #endif
> 5253 );
>
> Adding debug printf, I see that we have with clang:
> * sizeof(off_t) == 4
That seems to be wrong.
> * sizeof(long long) == 8
>
> Yet in src/auto/config.h, I see this (which is thus incorrect):
>
> /* Defined to the size of off_t */
> #define SIZEOF_OFF_T 8
>
> The configure script is getting SIZEOF_OFF_T incorrectly when using
> clang compiler.
>
> In src/auto/config.log, I see:
>
> configure:11536: checking size of off_t
> configure:11541: clang -o conftest -g -O2 -L/usr/local/lib
> conftest.c -lm -lncurses -lnsl -lselinux -lacl -lattr -lgpm >&5
> configure:11541: $? = 0
> configure:11541: ./conftest
> configure:11541: $? = 0
> configure:11556: result: 8
>
> Why do we compute size of 'off_t' in configure script, rather than
> simply use sizeof(off_t)?
That doesn't always work in #ifdef.
Perhaps the AC_SYS_LARGEFILE macro interferes? If it's not used
properly the sizeof(off_t) may differ depending on some flags.
> PS: You can install the clang compiler on Ubuntu with "sudo apt-get
> install clang".
> I had to add #include <wchar.h> in vim.h to be able to compile Vim
> successfully with clang.
Why? Hmm, perhaps it's needed in charset.c, because towlower and
towupper are used there. I'll do that. Does it work now?
--
I started out with nothing, and I still have most of it.
-- Michael Davis -- "Tonight Show"
/// Bram Moolenaar -- [email protected] -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ download, build and distribute -- http://www.A-A-P.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///
--
You received this message from the "vim_dev" maillist.
Do not top-post! Type your reply below the text you are replying to.
For more information, visit http://www.vim.org/maillist.php
