2008-05-29 15:42:25 Mark Phippard napisał(a):
> On Thu, May 29, 2008 at 9:34 AM, Philip Martin <philip_at_codematters.co.uk> wrote:
> > Arfrever Frehtes Taifersar Arahesis <arfrever.fta_at_gmail.com> writes:
> >
> >> 2008-05-19 20:24:13 Philip Martin napisa�(a):
> >>> arfrever_at_tigris.org writes:
> >>>
> >>> > + svn_error_clear(svn_dso_load(&dso, libname));
> >>> > + if (dso)
> >>> > + {
> >>> > + if (! apr_dso_sym(&provider_symbol, dso, funcname))
> >>> > + {
> >>> > + svn_auth_simple_provider_func_t func;
> >>> > + func = (svn_auth_simple_provider_func_t) provider_symbol;
> >>> > + func(provider, pool);
> >>> > + ret = TRUE;
> >>> > + }
> >>> > + }
> >>> > + return ret;
> >>> > +}
> >>> > +#endif
> >>>
> >>> I've seen this code before...
> >>>
> >>> Why is it acceptable to ignore errors from svn_dso_load?
> >>>
> >>> Why is it acceptable to use the dso variable if svn_dso_load returns
> >>> an error?
> >>
> >> Fixed in r31367.
> >
> > The code still drops the error. Why is this code different from the
> > RA loader code? Why don't you simply return the error?
>
> Is it because we expect the error to happen and we want to just know
> there was an error and fallback to the other mechanisms? For example
> if SVN is built with gnome-keyring and you are using the command line
> in a terminal where those libraries are not available?
Rather when Subversion is built with support for GNOME Keyring, but
GNOME Keyring isn't installed.
--
Arfrever Frehtes Taifersar Arahesis
Received on 2008-05-29 15:47:50 CEST