| Age | Commit message (Collapse) | Author |
|---|
|
|
|
relative to the initial input file pointer position, not
necessarily from the beginning of the file.
[!from_start]: Don't back up past the initial position of the
input file pointer.
(tail_lines): Call file_lines only if FD refers to a regular file
with its file pointer positioned at beginning of file. Otherwise,
call pipe_lines. This is a kludge. Once there's a decent test
suite, fix this properly.
Before, (echo 1; echo 2) > k; sh -c 'read x; tail' < k
would output both lines of the input file even though the first had
already been read. Reported by John Roll (john@panic.harvard.edu).
|
|
(md5_file, md5_check): Set it.
(main): Use it.
[OPENOPTS]: Depend explicitly on BINARY.
(md5_file): Take a new parameter, MD5_RESULT, and no longer
generate output.
(md5_check): Invoke md5_file instead of calling fopen directly.
When giving a diagnostic for a line with invalid format, also report
the line number.
(main): Generate output after a successful md5_file call.
[in many places]: Upon detection of an error, rather than exiting
immediately, issue a diagnostic, note that an error occurred and
exit later.
|
|
issue the usage message; also report why.
|
|
|
|
|
|
|
|
|
|
(main): Call the new function instead of doing all that in an if stmt.
Always use "%s" format rather than raw filename as format argument
in printf-style functions like error. Otherwise, filenames containing
e.g. `%s' lose.
Remove all traces of `old format'. There is only one format now.
The compatible one.
|
|
[__P]: Define macro.
|
|
indication of success rather than second argument.
|
|
(main): Report an error if md5_stream fails. Otherwise, running
`md5sum dir-on-mounted-filesystem' always reported the checksum
for an empty file. Now it gets the `is a directory' error.
Greg McGary reported that the released version got stuck in an
infinite loop with such arguments.
|
|
Properly handle file names with leading and trailing white space.
|
|
|
|
Reformat paragraph so line with new FSF address fits in 80 columns.
|
|
Fiddle comments.
|
|
|
|
|
|
|
|
|
|
|
|
(main): Rename local variable.
Add FIXME: allow newlines in filenames.
Detect and report failed fclose calls on stdout and checkfile_stream.
|
|
be of type size_t, not int.
(main): Declare NR and CHARS_READ to be of type long, not int.
|
|
(append_range): Declare FIRST and LAST parameters to be unsigned int,
not size_t.
(find_closing_delim): Clean up interface, separating boolean success
indicator and index.
(find_bracketed_repeat): Likewise.
(build_spec_list): Adapt to use new interfaces.
|
|
|
|
|
|
|
|
call to tail_lines. Otherwise, NetBSD lost because N_UNITS (of type
off_t, a 64-bit type on NetBSD) wasn't converted to the target type,
long. Reported by Dan Hagerty (hag@gnu.ai.it.edu).
|
|
|
|
|
|
omission of field spec only when character offset is specified.
Otherwise, -. +. would be accepted.
|
|
position specification. Omitting M is equivalent to specifying `0'
for M. From Rick Sladkey (jrs@world.std.com).
|
|
|
|
Before, running `printf 1234| ./od --width=4 --read-bytes=4'
printed output for 8 bytes -- the last four were garbage.
This happened only when the dump limit, N, was specified (with
--read-bytes=N) and N was a multiple of bytes_per_block
(usually 16, but 4 in this example). From Andreas Schwab.
|
|
|
|
|
|
|
|
|
|
(main): Write (de facto) standard Plumb/Lankester format.
New option: --quiet.
Check option no longer takes an argument.
When checking, exit status reflects success.
|
|
|
|
silently overflow, accepting a group id larger than INT_MAX.
(isnumber): Remove now-unused static function.
(change_file_group) [MAXUID]: Give a more descriptive message
when numeric group id is larger than MAXUID.
|
|
|
|
|
|
it away for compilers that don't grok it.
|
|
|
|
|
|
on systems.
|
|
conflict with max macro from some system's header files and bad pre-ANSI
compilers.
|
|
also check `unsigned long', as a concession to systems with
int's smaller than 32 bits.
|
|
|
