On 2021-11-25 11:51, Achim Gratz wrote:
Brian Inglis writes:
On 2021-11-24 12:54, Achim Gratz wrote:
Brian Inglis writes:
Problem mentioned by me some time ago, may be related to this, from
undocumented mandb handling of Windows localization catalog folders:
It isn't, which could have easily been tested.
Are you saying that the <CygwinRoot>/%SystemDrive% issue is unrelated
to mandb execution, contrary to the information from the OP?
Yes, all the time, and now again.
How can it be tested, as I would like to do so, as I get those symptoms.
mkdir -p /tmp/test/{s,p}d2 /tmp/test/{s,p}d3/sub
cd /tmp/test
env -i /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i SystemDrive=sd1 /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i ProgramData=pd1 /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i SystemDrive=sd2 /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i ProgramData=pd2 /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i SystemDrive=sd3/sub /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i ProgramData=pd3/sub /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i SystemDrive=sd3/nil /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i ProgramData=pd3/nil /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i SystemDrive=sd3/void/nil /usr/bin/cygstart --hide /usr/bin/dash -c echo
env -i ProgramData=pd3/void/nil /usr/bin/cygstart --hide /usr/bin/dash -c echo
ls -R
# cd ; rm -fr /tmp/test
The behaviour with absolute paths is left as an exercise for the reader.
What a mess! No sign of any obvious reports, except a decade old report
of the same bug from C#:
http://jeremymurray.org/doku.php/blog:20110804_directories_popping_up_named_systemdrive
blaming the use of ShellExecute, which may have changed in cygutils
cygstart, as the bug did not appear to occur until recently.
Patching setup to pass those env vars through to scripts would be one
approach.
Patching cygstart to set those env vars if unset, using
CSIDLs/KNOWNFOLDERIDs like cygpath -F is another.
I found that when setup is executed, as the PATH sanitized for scripts
includes cygpath -S, mandb appears to search directories in the
Windows PATH, for cat... directories, may remove the contents, even
when they are not manual pages, and may perform other poorly or
undocumented operations.
Again, in no variant of trying to reproduce this was I ever successful
in seeing that happen. So either this problem is no longer occuring or
there is some extra condition for it to trigger.
It appears that mandb climbs the tree trying to find related bin and
man dirs.
There's no code in mandb that obviously does this, so you'll have to
provide some evidence of when and where it happens. And it should
already show up in the output of 'manpath -dc' I think, since I've only
ever seen mandb removing something that is in catdirs. So for instance
if I drop bogus directories cat9 and de_AT into /var/cache/man (which
don't have a corresponding manpath obviously), I'll see:
--8<---------------cut here---------------start------------->8---
Removing obsolete cat directory /var/cache/man/cat9...
Seen mandir for /var/cache/man/ca; not deleting
Seen mandir for /var/cache/man/cs; not deleting
Seen mandir for /var/cache/man/da; not deleting
Seen mandir for /var/cache/man/da.ISO8859-1; not deleting
Seen mandir for /var/cache/man/da.UTF-8; not deleting
Seen mandir for /var/cache/man/de; not deleting
Seen mandir for /var/cache/man/de.ISO8859-1; not deleting
Seen mandir for /var/cache/man/de.UTF-8; not deleting
Removing obsolete cat directory /var/cache/man/de_AT...
--8<---------------cut here---------------end--------------->8---
Which is correct since /var/cache/man is the only catdir on my system.
If I drop these anywhere else in any of the directory in PATH mandb will
not touch them (in fact it never bothers to look there).
It may be having shared man and "stray" cat dirs for other Windows
package commands, cmd, Windows utilities, and systems (-m) under
/proc/cygdrive/c/usr/local/.
I don't know any other way I could keep the system level shared
contents accessible while isolating it.
Again, if that's the problem I have not been able to reproduce it.
I'll create a new 0p_man-db-patch.dash script to add -d after mandb
postinstall to produce diagnostic output, and get back.
I previously had a patch script to add around mandb in postinstall:
/usr/bin/nohup ... /usr/bin/mandb ... &
which did not create any artifacts I believe but can not be sure of
timing, and that could be an option for your script instead of cygstart?
--
Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada
This email may be disturbing to some readers as it contains
too much technical detail. Reader discretion is advised.
[Data in binary units and prefixes, physical quantities in SI.]
--
Problem reports: https://cygwin.com/problems.html
FAQ: https://cygwin.com/faq/
Documentation: https://cygwin.com/docs.html
Unsubscribe info: https://cygwin.com/ml/#unsubscribe-simple
