nixpkgs/pkgs/development/tools/profiling/pyflame/default.nix

66 lines
2.2 KiB
Nix
Raw Normal View History

2018-12-02 13:16:36 +01:00
{ stdenv, fetchFromGitHub, autoreconfHook, coreutils, pkgconfig
# pyflame needs one python version per ABI
# are currently supported
2018-12-02 13:16:36 +01:00
# * 2.6 or 2.7 for 2.x ABI
# * 3.4 or 3.5 for 3.{4,5} ABI
# * 3.6 for 3.6+ ABI
# if you want to disable support for some ABI, make the corresponding argument null
, python2, python35, python36, python3
2018-12-02 13:16:36 +01:00
}:
stdenv.mkDerivation rec {
pname = "pyflame";
version = "1.6.7";
2018-12-02 13:16:36 +01:00
src = fetchFromGitHub {
owner = "uber";
repo = "pyflame";
rev = "v${version}";
sha256 = "0hz1ryimh0w8zyxx4y8chcn54d6b02spflj5k9rcg26an2chkg2w";
};
nativeBuildInputs = [ autoreconfHook pkgconfig ];
buildInputs = [ python36 python2 python35 ];
postPatch = ''
patchShebangs .
# some tests will fail in the sandbox
substituteInPlace tests/test_end_to_end.py \
--replace 'skipif(IS_DOCKER' 'skipif(True'
# don't use patchShebangs here to be explicit about the python version
substituteInPlace utils/flame-chart-json \
--replace '#!usr/bin/env python' '#!${python3.interpreter}'
'';
postInstall = ''
install -D utils/flame-chart-json $out/bin/flame-chart-json
2018-12-02 13:16:36 +01:00
'';
doCheck = true;
# reproduces the logic of their test script, but without downloading pytest
# from the internet with pip
checkPhase = with stdenv.lib; concatMapStringsSep "\n" (python: ''
set -x
PYMAJORVERSION=${head (strings.stringToCharacters python.version)} \
PATH=${makeBinPath [ coreutils ]}\
PYTHONPATH= \
${python.pkgs.pytest}/bin/pytest tests/
set +x
'') (filter (x: x!=null) buildInputs);
meta = with stdenv.lib; {
description = "A ptracing profiler for Python ";
longDescription = ''
Pyflame is a high performance profiling tool that generates flame graphs for
Python. Pyflame uses the Linux ptrace(2) system call to collect profiling
information. It can take snapshots of the Python call stack without
explicit instrumentation, meaning you can profile a program without
modifying its source code.
'';
homepage = https://github.com/uber/pyflame;
license = licenses.asl20;
maintainers = [ maintainers.symphorien ];
# arm: https://github.com/uber/pyflame/issues/136
platforms = [ "i686-linux" "x86_64-linux" ];
};
}