Mail Archives: djgpp/2012/06/24/08:15:18
Hi,
On Sunday, June 24, 2012 2:39:42 AM UTC-5, Andris Pavenis wrote:
> On 23.6.2012 21:52, rugxulo AT gmail DOT com wrote:
> >
> > On Saturday, June 23, 2012 10:44:43 AM UTC-5, Andris Pavenis wrote:
> >> On 23.6.2012 10:19, Andris Pavenis wrote:
> >>> It seems that we need to have own version of gcc/ada/s-os_lib.adb
> >>> with modified temporary file name generation for DJGPP.
> >>>
> >>> Ada support is perhaps broken for non LFN systems before we have one.
> >>>
> >> You could try the test build (ada471b.zip only, other files should be OK
> >> from the earlier build)
> >> http://ap1.pp.fi/djgpp/gcc/4.7.1/v204/testonly/ada471b.zip
Oops, I forgot to use zipcmp (from OpenWatcom's old libzip), it would've made it easier to see what changed. Oh well, just my curiosity since the file is so big. ;-)
> >> Temporary file names should hopefully be compatible with non-LFN system
> >> however I have not tested it
> > Yes, it seems to work much better, though it still fails at the end saying that "s-stratt.ali is outdated and needs to be recompiled". I'm not sure if that's due to some internal hash mismatching, you forgetting to remake it, or if it's a timestamp issue (doubtful as I touch'd it and it didn't help).
> It should not be as I did full fresh bootstrap yesterday before
> putting files for testing.
>
> However looking at sources shows that the problem is LFN
> conflict mentioned in the announcement.
I had noticed when you originally mentioned this, but I didn't actually think it would bite anyone, especially me! :-O
> We have s-stratt.adb
> and s-stratt-xdr.adb that both resolves to the same
> non-LFN filename. They are simply 2 different implementation
> of the same stuff as far as I can understand from the comments
> in the source code. It is possible that unpacking binary archive
> mixed files and the result some hash does not match.
Apparently gnatbind is the one doing the complaining.
> Could You try on non-LFN system:
> - taking only s-stratt.adb and leaving s-stratt-xdr.adb out
> - taking only s-stratt-xdr.adb and leaving s-stratt.adb out
>
> Is one of these variants working OK
Yes, my bad, it builds, links, and runs fine with the GNAT (smaller, older) version of s-stratt.adb. I guess "unzip -qn" (no overwrite) is bad here and "unzip -qo" (overwrite) is good since the XDR version is unpacked first. ;-)
I have no idea what XDR and GARLIC are, heh, but presumably you can just rename the file to not conflict in 8.3 at all (or even delete it), esp. since we apparently don't need it!
P.S. If anybody else ever needs to build "Gautier's 3D Engine" for testing, the .ZIP with sources is here:
https://sites.google.com/site/rugxulo/eng3d018.zip?attredirects=0
- Raw text -