Re: A possible RA plugin ABI problem

Tobias Ringstrom <tobias@ringstrom.mine.nu> writes:
> The RA plugin init function takes an ABI version as it's first
> parameter to be able to verify that the caller's ABI version matches
> the plugin. The problem is that it is defined to return an svn_error_t
> pointer when the ABI does not match.
>
> If the new ABI would change the type of svn_error_t, the caller would
> most likely crash if it tried to show the error to the user.

If svn_error_t had to change, we would most likely add new fields on
the end, so it would still be binary compatible with old code.

So there is a problem here in some sense, yeah, but it seems quite
unlikely to cause trouble in practice, and IMHO is not worth worrying
about... ?

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Jan 13 17:15:46 2004