Re: [ft-devel] Large code changes without changing functionality

[Behdad Esfahbod]

On 13-03-15 03:18 AM, Werner LEMBERG wrote:
> 
>> Doesn't the c879f24562cc21bb062fb2faa86633a6c8c2371a change have API
>> implications?  Is the module error base visible to users?
> 
> Thanks for catching this, fixed now in git.  I doubt that many people
> use module errors, however.
> 
> The reason that I still support this is the possibility for more
> improvements: For example, instead of having `module errors', I could
> register each single source code file and assigning an ID to it.  If I
> now say that error codes are 32bit wide, the upper 16bit could hold
> the line number, the next 8 bits a file ID and the lowermost bits the
> real error code.  This would be a quite effective means to find out
> the source code origin of returned error codes even in non-debug
> builds.

It would, but I'd consider module errors an API mistake to begin with.  For
the past ten years everyone has been comparing FT error codes using good old
equality operator.  If no one uses a feature, there's no point developing it
further.  If anything, it should be removed.

The FT_THROW facility on the other hand is extremely useful.

behdad