Relocation does not relocate files which contain paths for dependent packages

[smuzaffar]

If I am not mistaken then https://github.com/bitsorg/bits/blob/main/bits_helpers/build_template.sh code i.e

{ grep -I -H -l -R "\(INSTALLROOT/$PKGHASH\|[@][@]PKGREVISION[@]$PKGHASH[@][@]\)" . || true; } | sed -e 's|^\./||' > "$INSTALLROOT/.original-unrelocated"

will only find those files which has INSTALLROOT/$PKGHASH in them. This will miss many other files which does not contain INSTALLROOT/$PKGHASH but still has build time path for its dependent packages e.g. root/bin/root-config has many paths for its dependent packages [a] but it was not relocated.

[a]

> cat /cvmfs/alice.cern.ch/el9-x86_64/Packages/ROOT/v6-36-04-alice2-2/bin/root-config | tr ' ' '\n' | grep /local/workspace/DailyBuilds/DailyO2Physics-slc9
configargs="FFTW_INCLUDE_DIR=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/FFTW3/v3.3.9-22/include
FFTW_LIBRARY=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/FFTW3/v3.3.9-22/lib/libfftw3.so
FREETYPE_INCLUDE_DIR_freetype2=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/FreeType/v2.10.1-20/include/freetype2
FREETYPE_INCLUDE_DIR_ft2build=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/FreeType/v2.10.1-20/include/freetype2
FREETYPE_LIBRARY_RELEASE=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/FreeType/v2.10.1-20/lib/libfreetype.so
FTGL_INCLUDE_DIR=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/SOURCES/ROOT/v6-36-04-alice2/v6-36-04-alice2/graf3d/ftgl/inc
GSL_CBLAS_LIBRARY=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/GSL/v2.8-4/lib/libgslcblas.so
GSL_CBLAS_LIBRARY_DEBUG=/local/workspace/DailyBuilds/DailyO2Physics-slc9/daily-tags.M3tUFd6X0g/slc9_x86-64/GSL/v2.8-4/lib/libgslcblas.so

[ktf]

Yes, indeed. As discussed the decision in alibuild was simply not to rely on a given extenal to provide paths to its dependencies, because in general it can be wrong, especially when downloading prebuilt packages. Notice this is in handled correctly when using CMake, so we never bothered beyond that. If you have an idea on how to properly handle those, I am more than happy to include it.

[pbuncic]

@smuzaffar, I think that @ktf is expecting input/code from you to address this.

[smuzaffar]

As I have mentioned in the face 2 face meeting, in order to have fully relocatable packages we need to make sure that the tools/configurations a package provide should also be properly relocated. I agree that bits/packaging system should not rely on the package to provide the dependency but there are build system/users tools which use the packages' tools/configurations to search for their dependencies paths. I would like to make sure that packages we build are usable by those build system otherwise we have to patch their build system.

The fix to relocate the package's tools/configuration should be trivial. All we need is to search for all text files with work directory path and relocated them. This should work for prebuild packages too. I will open a PR to do the proper relocation

[smuzaffar]

I have opened #69

[ktf]

I am traveling right now. I will merge it asap. Sent from Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: Malik Shahzad Muzaffar ***@***.***> Sent: Thursday, December 18, 2025 4:46:49 PM To: bitsorg/bits ***@***.***> Cc: Giulio Eulisse ***@***.***>; Mention ***@***.***> Subject: Re: [bitsorg/bits] Relocation does not relocate files which contain paths for dependent packages (Issue #36) [https://avatars.githubusercontent.com/u/4115138?s=20&v=4]smuzaffar left a comment (bitsorg/bits#36)<#36 (comment)> I have open #69<#69> — Reply to this email directly, view it on GitHub<#36 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAACSMGBQK3SNFVGTQ5Y63L4CLD6TAVCNFSM6AAAAACLN3PJAWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTMNZQHEZDMNBQGQ>. You are receiving this because you were mentioned.Message ID: ***@***.***>