exim 4.54-1 and broken symlinks

[Krzysztof Duleba]

Pierre A. Humblet wrote:

> I still don't understand how the .exe got to appear...
> What I observe here is that if a.exe exists, then
> ln -s a b produces b.exe ==> a.exe  (weird to me)
> However ln -s a.exe b produces b ==>a.exe
> So that explains how sendmail.exe appears when
> exim.exe exists, but doesn't explain how /bin/exim.exe
> appeared in the first place.

And ln -s -f a b produces b ==> a if a.exe doesn't exist yet.

> Anybody else has a symlink /usr/bin/exim.exe
> and/or /usr/sbin/sendmail.exe ?
> If so I will fix things in the postinstall of the next
> exim release, coming out soon.

I've checked some other computers and those symlinks don't have .exe
suffix there.

> Regarding your remark about setup, it should be OK to
> install a new exim-Y.exe, while the previous exim-X.exe is
> still running.  Setup will set the /bin/exim symlink to point
> to exim-Y.exe.
> exim-X will keep running until it re-execs following a kill -HUP,
> or until a reboot (assuming that cygrunsrv is using the
> /bin/exim symlink). Then you get exim-Y.exe.
> Doing it that way you can get your server to operate non-stop
> through an update. However I am not sure if the old exim-X.exe
> is going to be deleted.

In my case it wasn't. Anyway, I prefer to turn off a service for a
moment before an update rather than reboot afterwards.

> Also, each time you run exim-config, it checks if /bin/exim is
> executable. If not, it sets the symlink to the most recent version
> in  /bin/exim*.exe.

After the initial update (the unsuccessful one), even after rebooting,
/bin/exim.exe was a symlink to the old version of exim executable.
Reinstalling exim 4.54-1 (with exim turned off this time) fixed it.

> So, I don't see how forgetting to reboot after an upgrade and
> then running exim-config could mess up things. exim-config
> should see that exim is running and ask if you want to stop
> it. If you answer no, exim-X.exe keeps running. Otherwise
> it stops and next time you get exim-Y.exe In no case is the
> symlink affected.

Well, if it's not that then I don't have a clue. I guess we can forget
about it - I don't have time to dig into it and try to find an
explanation (or even to reproduce it) and it doesn't seem really
serious, too.

Thanks,
Krzysztof Duleba



--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/