On Tue, Apr 29, 2014 at 10:57:50 +0200, Martin Kletzander wrote: > On Tue, Apr 29, 2014 at 10:45:56AM +0200, Jiri Denemark wrote: > >On Tue, Apr 29, 2014 at 10:36:51 +0200, Michal Privoznik wrote: > >> Currently, there's an issue with virStrToLong_* APIs that they turn > >> "-1" into UINT_MAX. While this is not acceptable, it works on 64 bit > >> architectures and doesn't work on 32 bit ones. I know that much > >> cleaner solution is required, but given that we are in the freeze we > >> may as well just skip the test on 32 bits. > > > >I think a workaround using strtoull instead of stroul in virStrToLong_ui > >would be better, even though I don't like that solution either. Your > >workaround disables the check but keeps the functionality which is > >tested here broken. > > > > Moreover, Pavel found out that strtoul should set errno to ERANGE in > case of the problem described here. The only problem is that we *are* > checking for errno there and it just passes :( But only in case the non-negative value does not fit in. If it fits, no error is reported even though negative number is parsed. The API is just insane. Jirka -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list