Re: permissions (and other) problems

Jon Watte wrote:

>>I posted a kludgey patch to prevent that a few months ago;
>>it just ignored the SIGINT. It did this even on ra_dav, though, which
>>was probably overkill.
>>
>>
>
>A slightly less kludgey fix is to install a SIGINT handler which
>does a longjmp() to some routine which cleans up all currently
>registered "cleanup" handlers. Typically, something that starts
>a transaction will register transaction rollback as a cleanup task,
>and something which opens the database will register database close
>as a cleanup task...
>
No. We want to be portable, and doing anythint except setting a
sig_atomic_t global variable from within a signal handler is not portable.

The Berkeley DB utilities use a similar method to the one I proposed,
for the same reason, I assume.

>Note that Linux has problems fully restoring all CPU state when
>doing longjmp() out of signal handlers (the FPU control word is not
>correct) but that doesn't matter for this case :-)
>
>
That's exactly what I mean. Every system behaves slightly differently on
longjump, especially in a signal handler. Let's not play with fire if we
have a perefectly usable nuke handy. :-)

-- 
Brane Čibej   <brane_at_xbc.nu>   http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org