Re: GNUstep-corebase to be released next week

[Fred Kiefer]

For me (OpenSuse 64bit AMD) the only failing test is this one:

Failed test:     format.m:28 ... Float/Doubles are formatted correctly
expected 5.500000 5.5   3.000000e-06, but got 5,500000 5,5   3,000000e-06

This definitely is a problem of the locale, you should have a look at 
the base test in NSNumberFormatter/basic10_4.m on how to ensure a 
specific locale for the test.

Fred

On 18.07.2012 09:47, Sebastian Reitenbach wrote:
> On Wednesday, July 18, 2012 01:44 CEST, Stefan Bidi <stefanbidi@gmail.com> 
> wrote:
>
> > I believe I fixed the at least some of the failing tests.  I made
> > CFNumber test more verbose to get some insight in what is going on.
> > The CFString formating tests will be a little more tricky, this most
> > likely has to do with how ppc passes arguments.  Unfortunately, I have
> > no idea where to even start fixing this error.
>
> yep, I looked a bit into it, and it seems, this needs some byte swapping, 
> depending
> on CFByteOrderGetCurrent() or the like. I'll also look into it later today.
>
> So far, all, besides this timezone name now work for me on i386 and amd64.
> on macppc, only the following fails:
>
> Running Tests/CFNumber/general.m...
> Failed test:     general.m:24 ... SInt32 converted to SInt8 (251) correctly.
> Running Tests/CFString/format.m...
> Failed test:     format.m:14 ... Characters are formatted correctly
> expected r un, but got
> Failed test:     format.m:50 ... Strings are formatted correctly
> expected 0xc A longer than usual string., but got 0x0 A longer than usual 
> string.
> Failed test:     general.m:40 ... Standard localized name is correct.
> expected Central European Time, but got Central European Standard Time
>
> > The pthread thing should also be resolved, please give it another shot.
>
> Yes, works for me, as long as I don't have LIBS defined in configure 
> environment.
> I patched configure.ac again to fix that, a space is missing ;)
>
> But I've seen to detect the objc_getClass, you took the easy route.
> What if someone installs libobjc2 so that it must be linked with:
> -li-give-a-fuck-how-i-call-the-runtime
> then your test will fail again. Therefore I suggested to use input from
> gnustep-config --objc-libs
> I'm not good with this autoconf, but if you don't mind, I'll take a look at it 
> also.
>
> So the patch below fixes:
>   - the configure mentioned above
>   - a couple "no newline at the end of file" compiler warnings
>   - and in CFStringFormat.c, the warning:
>    CFStringFormat.c: In function '_CFStringAppendFormatAndArgumentsAux':
>    CFStringFormat.c:1057: warning: comparison is always false due to limited 
> range of data type
>
> but actually, I don't know why you don't put the CFUnknownType, or define it as 
> UINT16_MAX?
> then this weirdo casting I'm doing might not be necessary, which I doubt may 
> not even be right.
>
> cheers,
> Sebastian