Adding new library: gnustep-startup, which packages the core
libraries necessary for GNUstep: gnustep-make, gnustep-base,
gnustep-gui, gnustep-backend.
Fixes error: cannot coerce a partially applied built-in function to a string, at lib/sources.nix:59:32
I don't understand how this used to work, but doesn't work anymore?
lib.splitString was blowing up the stack in instances where the
.git/packed-refs file was too large. We now use a regexp over the
whole file instead.
Closes #16570
Similar to #14272, but fixes 4.5 build rather than generic.
- Ignores errors due to strict-overflow warnings
- Strips clang-only '-Wno-format-pedantic' flag out since this build
uses gcc
GoCD is an open source continuous delivery server specializing in advanced workflow
modeling and visualization. Update maintainers list to include swarren83. Update
module list to include gocd agent and server module. Update packages list to include
gocd agent and server package. Update version, revision and checksum for GoCD
release 16.5.0.
"This projects aims at gathering/developing new completion scripts that
are not available in Zsh yet. The scripts are meant to be contributed to
the Zsh project when stable enough."
Replace the implementation by seq and add a deprecation warning.
The semantics seems a little bit off, but the function should only be
used for debugging.
Quoting various characters that the shell *may* interpret specially is a
very fragile thing to do.
I've used something more robust all over the place in various Nix
expression I've written just because I didn't trust escapeShellArg.
Here is a proof of concept showing that I was indeed right in
distrusting escapeShellArg:
with import <nixpkgs> {};
let
payload = runCommand "payload" {} ''
# \x00 is not allowed for Nix strings, so let's begin at 1
for i in $(seq 1 255); do
echo -en "\\x$(printf %02x $i)"
done > "$out"
'';
escapers = with lib; {
current = escapeShellArg;
better = arg: let
backslashEscapes = stringToCharacters "\"\\ ';$`()|<>\r\t*[]&!~#";
search = backslashEscapes ++ [ "\n" ];
replace = map (c: "\\${c}") backslashEscapes ++ [ "'\n'" ];
in replaceStrings search replace (toString arg);
best = arg: "'${replaceStrings ["'"] ["'\\''"] (toString arg)}'";
};
testWith = escaper: let
escaped = escaper (builtins.readFile payload);
in runCommand "test" {} ''
if ! r="$(bash -c ${escapers.best "echo -nE ${escaped}"} 2> /dev/null)"
then
echo bash eval error > "$out"
exit 0
fi
if echo -n "$r" | cmp -s "${payload}"; then
echo success > "$out"
else
echo failed > "$out"
fi
'';
in runCommand "results" {} ''
echo "Test results:"
${lib.concatStrings (lib.mapAttrsToList (name: impl: ''
echo " ${name}: $(< "${testWith impl}")"
'') escapers)}
exit 1
''
The resulting output is the following:
Test results:
best: success
better: success
current: bash eval error
I did the "better" implementation just to illustrate that the method of
quoting only "harmful" characters results in madness in terms of
implementation and performance.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @edolstra, @zimbatm
