Stas Bekman <[EMAIL PROTECTED]> writes: > Joe Schaefer wrote: > > Stas Bekman <[EMAIL PROTECTED]> writes: > > > >>setaside API needs to be changed as it strays from the rest of the > >>API. Since all it can return is errors which are useless to trap it > >>shouldn't return anything. It should run MP_RUN_CROAK like other > >>methods do. It's still possible to trap those with APR::Error, but who > >> will. > > You should allow users to check for APR_ENOTIMPL, because > > But your implementation of modperl_bucket_sv_setaside doesn't ever return > it. I'm not following you, Joe.
apr_bucket_setaside is exposed for *any* type of bucket now, not just mod_perl buckets. Invoking it on other bucket types (like pipe and socket buckets currently) will return APR_ENOTIMPL. > > they might be able to handle that case the same way ap_save_brigade does. > > So I agree that it should throw an exception in void context, but > > would prefer if it didn't do that in scalar/list context. > > I don't think it makes sense in the particular case of mod_perl. The > only users of modperl buckets are modperl users, and setaside is now > implemented. So why > would users want to check any success codes explicitly? Someone writing a filter with mod_perl may need to setaside certain buckets which the filter did not itself create (ie those coming from the content handler in the output filter case, or those coming from mod_ssl in the input filter case). -- Joe Schaefer --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
