duan_jandClaude Fable 5
cb65827159
Installs panddas into its own prefix (--prefix + --ignore-installed) so the shared ccp4/9.0 module is never modified; PYTHONPATH prepend shadows the ccp4-bundled copy. ccp4/9.0 is build- and runtime dependency. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0152MuuYUaapP7CtxrDAfvBh
37 lines
1.3 KiB
Python
37 lines
1.3 KiB
Python
#!/usr/bin/env modbuild
|
|
|
|
pbuild::prep() {
|
|
:
|
|
}
|
|
|
|
pbuild::configure() {
|
|
:
|
|
}
|
|
|
|
pbuild::compile() {
|
|
:
|
|
}
|
|
|
|
pbuild::install() {
|
|
# ccp4/9.0 is a build_requires, so ccp4-python is on PATH here.
|
|
command -v ccp4-python >/dev/null || \
|
|
std::die 42 "PanDDA: ccp4-python not on PATH — ccp4/9.0 module did not load"
|
|
|
|
# Vendor doc says: uninstall bundled panddas, upgrade pip+numpy, pip install
|
|
# panddas — all of which would mutate the shared ccp4/9.0 module. Instead:
|
|
# --prefix keeps everything inside this module, --ignore-installed pulls
|
|
# fresh deps (incl. numpy) here rather than reusing/soiling ccp4's, and the
|
|
# modulefile's PYTHONPATH prepend shadows the ccp4-bundled panddas.
|
|
ccp4-python -m pip install --prefix="${PREFIX}" --ignore-installed \
|
|
"panddas==${V_PKG}" || \
|
|
std::die 42 "PanDDA: pip install failed — check network and ccp4-python version"
|
|
|
|
# stable alias so the modulefile needn't know ccp4's python X.Y
|
|
local -a sp=( "${PREFIX}"/lib/python*/site-packages )
|
|
[[ -d "${sp[0]}" ]] || std::die 42 "PanDDA: no site-packages under ${PREFIX}/lib"
|
|
ln -sfn "${sp[0]#"${PREFIX}"/}" "${PREFIX}/site-packages"
|
|
|
|
[[ -x "${PREFIX}/bin/pandda.analyse" ]] || \
|
|
std::die 42 "PanDDA: bin/pandda.analyse missing — install incomplete"
|
|
}
|