stow/TODO
2019-06-27 20:37:40 +01:00

73 lines
2.6 KiB
Plaintext

* Add support for pre/post-(un)install hooks
This would allow correct handling of the Info dir file via
install-info, amongst other things:
*** http://unix.stackexchange.com/questions/73426/dealing-with-gnu-stow-conflicts
*** http://article.gmane.org/gmane.comp.gnu.stow.general/6661
* Get permission for next documentation release to be under FDL 1.3
* Import a debian/ tree from an older package and update it.
* Import a .spec file from somewhere and update it.
* Distinguish between .stow and (undocumented) .nonstow / .notstowed
** .stow is for marking stow directories - avoids altering them
but also allows --override to work
** .nonstow should be only for protecting non-stow directories against modification by stow
but currently allows modification via --override
** .notstowed is only honoured by chkstow
** Documentation needs to be clear on this.
* Prevent folding of directories which contain ignored files
* Honour .no-stow-folding and --no-folding
* Add semi-automatic conflict resolution
(This idea is possibly obsoleted via --override and --adopt.)
*** STOW_RESOLVE_CONFLICTS="non_stow_symlinks=t stow_symlinks=r"
*** Add documentation about conflict resolution
* Autodetect "foreign" stow directories
From e-mail with meyering@na-net.ornl.gov:
> My /usr/local/info equivalent is a symlink to /share/info
> because I want installs on all systems to put info files in that
> directory. With that set-up, stow chokes on fact that
> /usr/local/info is a symlink.
[...] Stow is designed to be paranoid about modifying anything it
doesn't "own." If it finds a symlink in the target tree (e.g.,
/usr/local/info) which doesn't point into the stow tree, its
paranoid response is to leave it the hell alone. But I can see in
this case how traversing the link and populating the directory on
the far end would be OK. Question: is that a special
circumstance, or would it always be OK to populate the far end of
a symlink in the target tree (when the symlink points to a
directory in a context where a directory is needed)? And: if it's
a special circumstance requiring a command-line option, should the
option be a mere boolean (such as, "--traverse-target-links") or
should it be an enumeration of which links are OK to traverse
(such as, "--traversable='info man doc'")?
Does Version 2 fix this? (Kal)
I think that because it never needs to create /usr/local/info,
it only needs to check the ownership of links that it _operates_ on,
not on all the elements of the path.
* emacs local variables
Local Variables:
mode: org
End: