Normally, it is safe to assume two regular files are different when
their st_size values are different. However, that assumption may
be invalid if either value is zero, as happens with files on Linux
/proc and /sys file systems. Since skipping this optimization will
usually cost very little (one read syscall, to read zero bytes),
it is fine to accommodate those unusual files.
* src/analyze.c (diff_2_files): Do not assume regular files differ
just because their st_size values differ when one or more is 0.
* src/diff.c (compare_files): Likewise.
* tests/brief-vs-proc-stat-zero: New test.
* tests/Makefile.am: Add it.
* NEWS (Bug fixes): Describe it.
Reported by Stephan Müller in http://debbugs.gnu.org/21942
Problem reported by Vincent Lefevre in <http://bugs.gnu.org/16848>.
The simplest solution is to remove the TOO_EXPENSIVE heuristic
that I added to GNU diff in 1993. Although appropriate for
circa-1993 hardware, these days the heuristic seems to be more
trouble than it's worth.
* NEWS: Document this.
* doc/diffutils.texi (Overview): Modernize citations.
Remove mention of TOO_EXPENSIVE heuristic.
* src/analyze.c (diff_2_files): Adjust to TOO_EXPENSIVE-related
API changes in gnulib's diffseq module.
Problem reported by Vincent Lefevre in <http://bugs.gnu.org/16608>.
* NEWS:
* doc/diffutils.texi (Binary, Invoking diff): Document this.
* src/analyze.c (briefly_report): Return void, not int.
All uses changed. Do not futz with exit status. Simplify.
* tests/binary: Adjust to match new behavior.
Problem reported by Errembault Philippe in:
http://lists.gnu.org/archive/html/bug-diffutils/2013-03/msg00012.html
* NEWS: Document this.
* src/dir.c (compare_names): Fall back on file_name_cmp if
compare_collated returns 0, unless ignoring file name case.
(diff_dirs): Don't bother with the O(N**2) stuff unless ignoring
file name case.
* tests/Makefile.am (TESTS): Add strcoll-0-names.
* tests/strcoll-0-names: New file.
* NEWS (New feeatures): Update 3.3's news to mention --no-dereference.
Reported by Denis Excoffier.
* Makefile.am (old_NEWS_hash): Update, since this modifies old, and
normally-immutable NEWS.
* src/util.c (c_escape_char): New function.
(c_escape): New function.
(begin_output): Escape file names when needed.
* src/context.c (print_context_header): New names parameter.
(print_context_label): New name parameter.
* src/diff.h (print_context_header): Change prototype.
* tests/filename-quoting: New file.
* NEWS: Document this change.
diffutils is now designed to build with Cygwin or MinGW.
The old DJGPP stuff probably doesn't work anyway.
* Makefile.am (SUBDIRS): Remove ms.
* NEWS: Document this.
* configure.ac (AC_CONFIG_FILES): Remove ms/Makefile.
* ms/Makefile.am, ms/README, ms/config.bat, ms/config.sed:
* ms/config.site: Remove.
* NEWS: Document this.
* doc/diffutils.texi (Comparing Directories): Likewise.
Also, document that these options work at the top level.
* src/diff.c (compare_files): Treat EBADF like ENOENT, to handle
the case where "-" is closed. Allow the other file to be
STDIN_FILENO, in case it's "-".
* tests/Makefile.am (TESTS): Add new-file.
* tests/new-file: New file.
Derived from Tim Waugh's patch in:
http://lists.gnu.org/archive/html/bug-diffutils/2011-08/msg00034.html
* NEWS, doc/diffutils.texi (diff Options): Document this.
* src/diff.c (compare_files): Implement this, by using
find_dir_file_pathname.
* src/diff.h (find_dir_file_pathname): New decl.
* src/dir.c: Include filenamecat.h.
(find_dir_file_pathname): New function.
Problem reported by Christoph Anton Mitterer in:
http://lists.gnu.org/archive/html/bug-diffutils/2010-08/msg00000.html
* NEWS: Mention this bug fix.
* src/dir.c (compare_names_for_qsort): Fall back on file_name_cmp
if two distinct entries in the same directory compare equal.
(diff_dirs): Prefer a file_name_cmp match when available.
* tests/Makefile.am (TESTS): New test colliding-file-names.
* tests/colliding-file-names: New file.
and the final hunk plus context-length aligns exactly with the end
of the newline-lacking file. Diff would fail to output the required
"\ No newline at end of file" line, thus rendering the output invalid.
This bug appears to have been introduced by 2006-05-07
commit 58d0483b, "(find_identical_ends): Fix huge performance bug...",
at least to the extent that reverting that change fixes the bug.
Considering the stated effect of that change and lack of metrics,
reverting it is not an option, so here we take a more direct approach.
Given these inputs,
printf '\n1'>a; printf '\n0\n\n1'>b
and running diff like this:
./diff -U1 a b
for input file "b", the pointer, files[1].linbuf[4][-1], to
the last byte on the final line was mistakenly pointing at the
sentinel newline at EOF, rather than at the preceding byte.
(gdb) p files[1].linbuf[4][-1]
$3 = 10 '\n'
Thus, this test in the final print_1_line call:
if ((!line_flag || line_flag[0]) && limit[-1] != '\n')
fprintf (out, "\n\\ %s\n", _("No newline at end of file"));
would fail, because limit[-1] (which is files[1].linbuf[4][-1])
was mistakenly '\n', rather than the desired '1'.
My first thought was simply to adjust the final linbuf[line] setting,
at the end of io.c's find_and_hash_each_line function function:
if (p == bufend)
- break;
+ {
+ if (current->missing_newline)
+ --linbuf[line];
+ break;
+ }
But that would make diff misbehave with this input
(same as above, but with a newline appended to "a"),
printf '\n1\n'>a; printf '\n0\n\n1'>b
./diff -U1 a b
due to the block (100 lines above) that is triggered in that case
(but not in the both-files-missing-newline case):
if (p == bufend
&& current->missing_newline
&& ROBUST_OUTPUT_STYLE (output_style))
{
/* This line is incomplete. If this is significant,
put the line into buckets[-1]. */
if (ignore_white_space < IGNORE_SPACE_CHANGE)
bucket = &buckets[-1];
/* Omit the inserted newline when computing linbuf later. */
p--;
bufend = suffix_begin = p;
}
Note how "p" is decremented and "bufend" adjusted.
When that happens, we certainly don't want to decrement
"bufend" yet again.
Since there is no other way to determine at the end whether "bufend"
was already decremented, add a new variable to serve as witness.
* NEWS (Bug fixes): Mention it.
Reported by Timo Juhani Lindfors in http://bugs.debian.org/577832.
Forwarded by Santiago Vila.
* src/diff.c (print_context_function): For languages like Ada
that allow local functions and procedures, the plain context
function line may start with enough blank characters that the
function name does not get completely printed in the 40
characters limit. This patch solves this problem by removing
these useless initial blank characters.
* NEWS (Changes in behavior): Mention this change.
With this change, "printf '\0'|diff - /dev/null" now prints
"Binary files - and /dev/null differ" once again.
This reverts 2002-06-28 commit a352f0980,
"(briefly_report): Don't say "Binary files differ", ...".
* src/analyze.c (briefly_report): Do include the "Binary " prefix
in the diagnostic, when !brief.
* NEWS (Changes in behavior): Mention the diagnostic change.
Reported by Andreas Hoenen in http://bugs.debian.org/570064.
* src/Makefile.am (paths.h): Emit a definition of DEFAULT_DIFF_PROGRAM
that is simply "diff" (or whatever $(transform) would convert that to,
e.g., "gdiff"). This makes it so that tests can work without first
installing diff, and so that the binaries do not hard-code $(prefix).
* NEWS (Changes in behavior): Mention this.
* NEWS: s/the the/the/.
Adjust style to fit what some of the hooks and release tools expect.
* cfg.mk (local-checks-to-skip): Remove sc_the_the.
(old_NEWS_hash): Update.
(specify_ignore_initial): Reword to avoid gcc -W warnings.
(main): Use freopen instead of setmode, since freopen is in POSIX.
* src/context.c: Revert most 2004-09-01 changes. Then:
(TIMESPEC_NS): Remove. All uses replaced by
get_stat_mtime_ns.
Include stat-time.h, strftime.h.
(nstrtime): Remove decl.
* src/diff.c: Revert most 2004-09-01 changes. Then:
Don't include <posixver.h>, <quotesys.h>, <setmode.h>.
Include <sh-quote.h>, <stat-time.h>, <timespec.h>.
All uses of quotesys replaced by sh-quote.
(main, compare_files):
Use freopen instead of setmode, since freopen is in POSIX.
(main): Don't complain about "diff -NUM'.
(main, set_mtime_to_now):
Adjust to stat-time.h macros when accessing nanoseconds.
* src/diff3.c: Include sh-quote.h rather than quotesys. All uses
changed.
* src/dir.c (dir_read): excluded_filename renamed to
excluded_file_name.
* src/io.c: Don't include <setmode.h>.
(sip, read_files): Remove binary file stuff, leaving a FIXME behind.
A DOS expert needs to look at this.
* src/diff.c: Include sh-quote.h rather than quotesys.h.
All uses changed.
* src/system.h: Include verify.h.
(verify): Remove. All uses changed to verify.h version.
Include <intprops.h>.
(TYPE_SIGNED, TYPE_MINIMUM, TYPE_MAXIMUM): Remove. Now uses
intprops.h versions.
(O_BINARY): New defns, taken from coreutils.
* src/util.c: Include sh-quote.h rather than quotesys.h.
All uses changed.
