Re: gorm loading problem / binary compatibility

[Riccardo Mottola]

Hi,

Fred Kiefer wrote:
> Thank you for pointing out this issue. It looks like different 
> compilers treat enumerators differently. Some seem to turn them into 
> unsigned int, while others use int. To prevent this sort of problems I 
> went through the whole gui encoding code and replaced enumerator types 
> with the basic types that we used before. The should solve the problem 
> for the future.
Indeed. Apparently this happened with gcc 4.7.2 vs an older gcc 4.x version
> For your specific gorm file it would be best to recreate it from 
> scratch or from an older backup. If this is not possible you could try 
> to apply only half the change I made for NSParagraphStyle, the 
> encoding bit and leave the decoding as it was. That way you are able 
> to load the file in the incorrect format and store it in the corrected 
> one. But don't forget to switch to the full change after that, or you 
> wont be able to load the file again.
With your change I am able to load the Gorm file actually, so it gets 
decoded properly, it is the NetBSD machine that needs the fix. I'll 
experiment around and let you know.
Sadly, there is no backup or, worse actually, the back up I have was 
already not working.
> This also means that we need to do another gui release very soon, to 
> prevent others from creating incorrect gorm files. (Although I don't 
> expect to many people will store paragraph styles or events in gorm 
> files) And that we need a lot more testing code for gui. If we had 
> code that tries to load pre-created archives for every class that 
> supports encoding, we would have noticed earlier.
Actually, I did nothing special. Could it just be that this paragraph 
style gets encoded when you alight left or right a NSTextField used as 
label? This would be reasonably common.

This is a bug that could be only caught when playing with different 
computers and compiler versions, on a single machine everything was 
clean and consistent.

A release would be good probably. I'll email you privately a better test 
case for the table cells, so that perhaps we can finalize that too 
before a minor update release, which is going to happen for base too!

Riccardo