Re: [PATCH] Snarf-documentation was working by accident

[Daniel Colascione]

Thanks for taking a look at the problem.

On 8/13/10 3:35 AM, Jan Djärv wrote:
> It is not for the size, it is the fact that some symbols are defined
> several times, like x-popup-dialog for X, W32 and NS.  We don't want the
> NS help string in an X build.

That only applies to the C code though, right? There's also quite a bit
of logic surrounding delay-loading docstrings from *regular* compiled
elisp files. I just can't but wonder whether all that complexity is
still needed.

> I fixed the issue in a different way.

I don't think your fix does the right thing here.

The grammar of DOC is roughly:

DOC = FILE-ENTRY+
FILE-ENTRY = FILE-HEADER (VAR-ENTRY | FUNCTION-ENTRY)*
FILE-HEADER = "^_S" FILE-NAME "\n"
VAR-ENTRY = "^_V" SYMBOL-NAME "\n" DOCSTRING
FUNCTION-ENTRY = "^_F" SYMBOL-NAME "\n" DOCSTRING

Your patch resets skip_file each time we see a "^_", but skip_file is
supposed to apply to an entire file-entry. Consequently, if I'm reading
your patch correctly, skip_file is always 0 when we encounter a
VAR-ENTRY or FUNCTION-ENTRY, which will cause other problems.

(Also, the comment on line 643 is wrong and doesn't talk about the
FILE-HEADER case.)