1) no need to include sys/stat.h
2) remove the enum which just added a layer too thick on this simple
program
3) argc-style, other style
4) weprintf instead of enprintf, then save the error-message of
execvp before and return the proper status.
5) write consistent "not reached" comment.
col is used to display troff documents in ttys, removing the reverse
line feeds generated by .2C in ms. This implementation keeps the limit
of 256 lines of 800 characteres of the original implementation.
1) Refactor the manpage, which has been a bloody mess, documenting
fantasy-flags (-d for example) and add a STANDARDS section
2) fix usage()
3) sort ARG-block
4) Check return-value of stat() separately, so a lack of permissions
doesn't tell the user "the directory doesn't exist", which could
be a bit confusing.
5) Add empty line before return.
1) Fix usage()
2) Group local variables
3) Idiomatic argv-loop
4) BUGFIX: When the m-flag is specified, POSIX clearly says:
"Set the file permission bits of the newly-created FIFO to the specified mode
value."
This means, that if mkfifo() fails for some reason, it should not try to
chmod() the given path (which has been fixed with the "else if")
A simple testcase is:
$ touch testfile; mkfifo -m 000 testfile;
GNU mkfifo(1): ls -l testfile
-rw-r--r-- 1 testfile
sbase mkfifo(1): ls -l testfile
---------- 1 testfile
5) Add blank line before return
1) val is sufficient as "int" (read the standard)
2) BUGFIX: If getpriority fails, it returns -1 and sets errno.
Previously, it would correctly catch the errno but not take
care of the fact that by then val has been decremented by 1.
Only change val if the getpriority-call has been successful.
3) Add LIMIT()-macro from st to increase readability.
4) setpriority returns < 0 on failure
5) Remove bikeshedding-comment. Read the standard if you wonder.
6) return-value trick from env(1)
1) style fix (don't arrange local variables)
2) BUGFIX: Previously, if ret was turned 1 for some folder, it
would disable the p-flag for every following folders, which
is not desired.
Instead, the "else if" makes sure that the p-flag-section is
only entered when the initial rmdir succeeds.
3) BUGFIX: Previously, the program would cancel with eprintf if
it failed to remove one folder in the parent-pathname.
This is not desired, as we have other folders pending.
Instead, print a warning for the current failing parent-folder,
set ret to 1 and break off the loop at this point.
This allows to finish the other pending folders without issues.
1) Update usage as already done in the manpage
2) group and sort local variable declarations
3) Be pedantic about the number of options. Don't just ignore it
if argc > 1.
1) Shorten synopsis and reflect this in the manual
2) Use argv0 in usage()
3) Decrement argc in argv-loop for consistency
4) Make it clearer which error-code results from which errno in enprintf
5) Use idiomatic for-loop also for environ. Don't increment these pointers
in the loop itself!
1) Fix usage ... spacing
2) use *argv instead of argv[0] in the idiomatic for-loop
3) Stop the naïve usage of "/dev/fd/0" and use plain stdin
instead (This also makes error-messages more consistent).
4) Add newline before return
5) Remove comma in manpage
Now that -c behaves correctly, the tools are pretty much done.
Only the manpages were not clear enough what happens when you
specify the c-flag.
This is fixed now.
1) be stricter which number of arguments is accepted (1 or 2)
2) basename already returns a pointer to "." is argv[0] is ""
3) No need to check for *p != '/', because basename() only returns
a string beginning with '/' which has length 1, so if strlen(p)
== 1, the only way for suffix to be "evaluated" is for off to
be > 0, being equal to suffix being "", but "" != "/".
4) don't calculate strlen twice for each string. Store it in a
ssize_t and check if it's > 0.
Okay, so why another section?
The finished-section applies to general feature-completeness and
manual status. It somehow is not an indicator for general code-
clarity, so the audited-column reflects a thorough audit of the
underlying code and optimization.
Take a look at the upcoming basename(1)-patch for an example on
how this goes.
Previously, the string-length was limited to BUFSIZ, which is an
obvious deficiency.
Now the buffer only needs to be as long as the user specifies the
minimal string length.
I added UTF-8-support, because that's how POSIX wants it and there
are cases where you need this. It doesn't add ELF-barf compared to
the previous implementation.
The t-flag is also pretty important for POSIX-compliance, so I added
it.
The only trouble previously was the a-flag, but given that POSIX
leaves undefined what the a-flag actually does, we set it as default
and don't care about parsing ELF-headers, which has already
turned out to be a security issue in GNU coreutils[0].
[0]: http://lcamtuf.blogspot.ro/2014/10/psa-dont-run-strings-on-untrusted-files.html
This is a particularly interesting program.
I managed to implement everything according to POSIX except how
octal escapes are specified in the standard, which is yet another
format compared to the one demanded for tr(1).
This not only confuses people, it also adds unnecessary cruft
for no real gain.
So in order to be able to use unescape() easily and for consistency,
I used our initial format \o[oo] instead of \0[ooo].
Marked as optional is UTF-8 support for %c in the POSIX specification.
Given how well-developed libutf has become, doing this here was more
or less trivial, putting us yet again ahead of the competition.
and mark it as finished in the README.
Specifically, add a small section on the compression flags, which
are basically an infected GNU limb which should be removed from
the face of the earth as soon as possible.
The algorithm had some areas which had potential for improvement.
This should make cmp(1) faster.
There have been changes to behaviour as well:
1) If argv[0] and argv[1] are the same, cmp(1) returns Same.
2) POSIX specifies the format of the difference-message to be:
"%s %s differ: char %d, line %d\n", file1, file2,
<byte number>, <line number>
However, as cmp(1) operates on bytes, not characters, I changed
it to
"%s %s differ: byte %d, line %d\n", file1, file2,
<byte number>, <line number>
This is one example where the standard just keeps the old format
for backwards-compatibility. As this is harmful, this change
makes sense in the sense of consistentcy (and because we take
the difference of char and byte very seriously in sbase, as
opposed to GNU coreutils).
The manpage has been annotated, reflecting the second change, and
sections shortened where possible.
Thus I marked cmp(1) as finished in README.
Use size_t for all counts, fix the manpage and refactor the code.
Here's yet another place where GNU coreutils fail:
sbase:
$ echo "GNU/Turd sucks" | wc -cm
15
coreutils:
$ echo "GNU/Turd sucks" | wc -cm
15 15
Take a bloody guess which behaviour is correct[0].
[0]: http://pubs.opengroup.org/onlinepubs/009604499/utilities/wc.html
and mark it as finished in the README.
Previously, it would only parse octal mode strings. Given
we have the parsemode()-function in util.h anyway, why not
also use it?
and mark it as finished in the README.
This is another example showing how broken the GNU coreutils are:
$ echo -e "äää\tüüü\tööö" | gnu-expand -t "5,10,20"
äää üüü ööö
$ echo -e "äää\tüüü\tööö" | sbase-expand -t "5,10,20"
äää üüü ööö
This is due to the fact that they are still not UTF8-aware and
actually see "ä" as two single characters, expanding the "äää" with
4 spaces to a tab of length 10.
The correct way however is to expand the "äää" with 2 spaces to a
tab of length 5.
One can only imagine how this silently breaks a lot of code around
the world.
WHAT WERE THEY THINKING?
which we are not planning to include into sbase.
What's left to discuss is how we're going to handle them in the
tools (dump usage() or silently ignore them).
Now you can specify a multibyte-delimiter to cut, which should
definitely be possible for the end-user (Fuck POSIX).
Looking at GNU/coreutils' cut(1)[0], which basically ignores the difference
between characters and bytes, the -n-option and which is bloated as hell,
one has to wonder why they are still default. This is insane!
Things like this personally keep me motivated to make sbase better
every day.
[0]: http://git.savannah.gnu.org/gitweb/?p=coreutils.git;a=blob;f=src/cut.c;hb=HEAD
NSFW! You have been warned.
One major milestone is to have the sbase-tools supporting UTF-8.
Tools like cut(1) with the -n flag don't make sense otherwise.
And while the gnu coreutils cut(1) blatantly ignores such an
important aspect, we will not tolerate this madness and mark it
as a TODO in the main README.
Since most tools inherently support UTF-8 anyway, this just concerns
tools which mangle with text or search in it in special ways.
and mark it as finished in README.
One small rationale on the way the manpage is set up: Looking at
the coreutils manpage, it does not invite to be a quick reference
guide, whereas I wrote this manpage to be short and concise in regard
to the information the advanced user needs.
No one needs to explain what an octal number is. That's not part of
the scope of this manpage.
Also, nobody wants to read a block of text just to find out how
to build an octal mode string.
to mark tools considered finished.
Finished doesn't mean work has stopped on these, but that these
programs are in a satisfying state according to the current suckless
coding practices, this includes having a
1) mandoc manpage
2) clean code
In most cases, 1) was the failing criterion. So in the interest of
finishing more tools and if you want to, well-written mandoc man-
pages are very much appreciated.
Get rid of it for now as it is not really widely used. We can do
a simple implementation when time comes.
Remove the table from README because it is not easy to edit unless
you use emacs.
We seem to have problems building individual tools across various
make implementations. If anyone can step up and fix this we will
remove the dependency on GNU make.
