Re: [1/3] scrrun: Implement filesys_DriveExists.

[Joachim Priesner]

Thanks for having a look at the code, will fix the method wrapper issue and resend later.
> Not necessary wrong, but it'd be better if all that validation was
> GetDriveTypeW responsibility. I don't see many tests for it in
> kernel32/tests, could you please add some, so we can get rid of this
> extra logic?

GetDriveTypeW does not accept strings consisting of only a single drive letter, but filesys_DriveExists does. So we cannot rely on GetDriveTypeW's validation and have to do our own.

I'll send the GetDriveType tests in a separate patch, good to have them anyway :)
> I haven't looked much at its implementation but it feels 
> like DRIVE_UNKNOWN would also qualify as non-existent drive.

The Windows implementation of DriveExists returns true for DRIVE_UNKNOWN. (I tested using a native scrrun.dll by patching GetDriveTypeW to always return DRIVE_UNKNOWN.)
> > -
> >    DriveSpec[0] = toupperW(DriveSpec[0]);
> > -
> >    if (DriveSpec[0] < 'A' || DriveSpec[0] > 'Z'
> > -
> >            || (len >= 2 && DriveSpec[1] != ':')
> > -
> >            || (len == 3 && DriveSpec[2] != '\\'))
> > -
> >        return E_INVALIDARG;
> > -
> >    hr = filesys_DriveExists(iface, DriveSpec, &drive_exists);
>
> I'm not sure about that sanity check, the reason is that DriveExists
> doesn't fail on invalid drivespec?

Correct, we cannot distinguish between a nonexisting drive and a malformed parameter just by looking at hr.

Joachim