| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
This prevents it getting upset if the sorting for the ignore list used a
different collation.
|
| |
|
|
|
|
|
| |
I thought shell script stripped \r from the end of variables, but that
does not actually seem to be the case. I think it's just newlines.
|
|
|
|
| |
Required for the mktd(1df) boilerplate.
|
|
|
|
| |
Way better, and more generally useful.
|
|
|
|
|
| |
Don't report to the user when a repository is created, and ignore errors
from git-diff-index(1) for when there is no HEAD
|
|
|
|
|
|
|
|
| |
ShellCheck:
In bin/clog.sh line 2:
self=clog
^-- SC2034: self appears unused. Verify it or export it.
|
|
|
|
|
|
|
|
| |
ShellCheck 0.4.7 pointed to this indirectly:
In bin/sshi line 24:
printf "%s:%u -> %s:%u (%s)\n" \
^-- SC1117: Backslash is literal in "\n". Prefer explicit escaping: "\\n".
|
|
|
|
|
|
|
|
|
|
| |
ShellCheck 0.4.7 was upset about this:
In bin/sra line 7:
ssh -qt -- "$hostname" "$@" <&3 # shellcheck disable=SC2029
^-- SC1073: Couldn't parse this simple command.
^-- SC1126: Place shellcheck directives before commands, not after.
^-- SC1072: Fix any mentioned problems and try again.
|
|
|
|
|
| |
This makes a bit clearer how awkward the rlwrap(1) code is, too. It may
not be worth keeping it.
|
|
|
|
| |
This is shorter and tidier.
|
|
|
|
|
| |
This makes it a little more flexible, if you genuinely don't want a
newline in the output.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The format in the first argument does not need to be evaluated, so it
can be passed in a simple awk variable.
The second argument is evaluated, by design, so code injection is
trivial. It's probably a good idea to warn users about this explicitly.
$ ax '0);system("cat /etc/passwd")'
Make the whole thing a little terser, too, with the awk program
construction, variable assignment, and invocation all on one line.
|
| |
|
|
|
|
|
|
|
|
|
| |
I realised I could make this work by recording a single byte in the
temporary file with dd(1) and then emitting that and then the rest of
the input with cat(1) if the file ended up with a byte in it.
This lets me remove the CAVEATS section from the manual, as it no longer
applies.
|
|\
| |
| |
| |
| | |
* feature/d2u-u2d-consist:
Make d2u(1df)/u2d(1df) like their stream analogues
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Remove the idempotency guarantee, and simplify the ed(1) scripts. See
commits 2905980 and cd8e9cc:
>commit 29059804f7708413843687c1764bc845d374a82d
>Author: Tom Ryder <tom@sanctum.geek.nz>
>Date: Fri Nov 3 13:58:23 2017
>
> Remove idempotency assert for sd2u(1df)/su2d(1df)
>
> Commit cd8e9cc applies a cleaner implementation of these tools but
> loses the idempotency:
>
> * Repeated su2d applications will result in double \r, so \r\r\n
> * Repeated s2ru applications will result in an extra newline at the
> end of the file, because the whole file will be interpreted as
> one line
>
> However, I am OK with this, as I think of the operation as simpler
> and more predictable, and I wouldn't apply it as a means to "force"
> a file of unknown or various line-ending types to one type.
>
>commit cd8e9cc27f7dd9d360b64f4a34b8c2d048f42e45
>Author: Tom Ryder <tom@sanctum.geek.nz>
>Date: Fri Nov 3 13:46:30 2017
>
> Apply simpler method for sd2u(1df) and su2d(1df)
>
> This method is shorter, easier to read, and more idiomatic.
|
|/
|
|
| |
This method is shorter, easier to read, and more idiomatic.
|
|
|
|
|
| |
If we don't do this, we run the risk of expanding an arbitrary macro
VERSION on any given system.
|
|
|
|
|
|
|
|
|
|
| |
This is mostly just for fun, but could be handy later on when I'm
playing with distributed or automated deployments of tagged and verified
releases.
Like a few of the other shell scripts, this is built by abusing my
mi5(1df) wrapper to get static details baked into the shell script that
are only known at runtime.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Modern OpenSSL builds, at least on Debian Stretch, don't even include
this as an option anymore, because SSLv2 has been so broken for so long.
There's not really much point trying to keep it.
$ openssl version
OpenSSL 1.1.0f 25 May 2017
$ openssl -no_ssl2
Invalid command '-no_ssl2'; type "help" for a list.
|
|
|
|
| |
This is more idiomatic, and explicitly initialises the result variable.
|
| |
|
| |
|
|
|
|
| |
Because I forgot
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
Forgot to remove the square brackets
|
|
|
|
| |
Debian's ancient mawk doesn't support them
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Specific values for these tasks get chosen way more often than other in
mawk, and it seems to be caused by the random seed being above a certain
value. Not sure if it's a bug or how it interacts with the POSIX
standard, but this seems to fix it.
|
| |
|
|
|
|
| |
Mostly inspired by suggestions from gawk --lint
|
| |
|
|
|
|
|
|
| |
gawk's --lint option complains:
> awk: .dotfiles/bin/trs.awk:7: warning: null string for `FS' is a gawk extension
|
|
|
|
|
| |
I forgot that Debian's awk(1) is still a mawk that doesn't implement
e.g. [:alpha:]
|
| |
|
| |
|
|
|
|
|
| |
Removes the need for the temporary file. Also refactor pks(6df) to
accommodate it.
|
| |
|
| |
|
| |
|
| |
|