Re: Compiled files without sources????

[David Engster]

David Kastrup writes:
> David Engster <address@hidden> writes:
>> The compromise was to let grammar development happen in CEDET
>> upstream. I did suggest this in good faith; in fact, it would have taken
>> almost no work at all to commit the plain grammar files to Emacs
>> trunk. But without Bovine and Wisent, the CEDET parser generators, those
>> grammar files are practically useless; contrary to what was written on
>> LWN/Slashdot/wherever, the grammars are not Bison compatible. To do any
>> actual work with them, you need to get CEDET, and then you will get all
>> the grammars anyway.
>
> Not the corresponding versions to what is included in Emacs.

True. I was fixed on development, and so you would always work with the
newest grammars form bzr (there are very few changes there anyway). The
older ones are of course not lost, but we should have included them. I
understand that. AFAIK the grammars used in Emacs are those from the
CEDET 1.0 release.

> Now Chong stated that he had to hand-edit the resulting grammars to make
> them fit into what is included in Emacs.  That seems like a bad idea
> with regard to meeting the source obligations and maintainability.

Yes. As you know, this was meant as a transitory period. I know this
sounds silly after two years, but it's already much work just keeping
CEDET working with Emacs24 development and fixing bugs which pop up on
the mailing lists. And sometimes I also like to actually code something
new.

> So what would be entailed to let upstream CEDET generate compiled
> grammars directly usable in Emacs?  Why don't they work out of the box
> when the parser runtime is ostensibly what is present in upstream
> CEDET?  Can the differences be factored out into variables?

The most important change is the 8+3 file renaming; this is replicated
in the file-rename branch. Also, the CEDET compilation process was
complicated and dependant on autoloads from EIEIO classes and methods,
which Emacs doesn't support and which therefore was also removed; Lluís
has written a completely new Makefile for that. I haven't tested yet if
the grammars generated in the new branch work right away, but we should
be close. I'm pretty sure there aren't any differences regarding the
actual grammar rules.

-David