bug#25071: 26.0.50; Various Forth related improvements for etags

[Helmut Eller <eller.helmut@gmail.com>]

On Wed, Nov 30 2016, Eli Zaretskii wrote:

>> With the patch, etags creates explicit tag names.  That works better in
>> combination with tag-exact-match-p.
>
> Can you explain more about this?  Explicit tag names bloat the TAGS
> files, so other languages don't use that by default, except where
> really needed.

Yes, the TAGS file becomes at most twice as big.  That's a minor
disadvantage that I'm happy to trade for even tiny improvements in
precision.

I also note that, Gforth (the GNU Forth implementation) ships a TAGS
file for the built-in definitions with explicit tag names (usually
installed in /usr/share/gforth/<version>/TAGS).  They use the Gforth
compiler, not etags, to create this TAGS file.  Obviously they aren't
bothered by the bloat either.

> Could it be that only some tags need to be written as
> explicit, e.g. because they contain some special characters?

Yes, that's the main reason.  E.g. with the old version a source file
test.fth with this content:

: (foo) 123 ;

produced this TAGS file:
\f
test.fth,17
: (foo)\x7f(foo.1,0

Hmm, it already created explicit tag names before.  But with "(foo" as
name, which is wrong.  The name should be "(foo)".

In case your are not familiar with Forth syntax: names in Forth are
separated by whitespace and can contain "(", ")", "," or other
non-whitespace characters.  A word in parentheses like "(foo)" is a
widely used naming convention for internal/auxiliary definitions.

The old version of etags.c used get_tag which uses notinname to find the
end of the name, but that excludes ")", which is wrong for Forth.

> Also, it sounds like you send a patch wrt emacs-25 branch; if so,
> please rebase on master, since that is where the patch will be pushed
> eventually.  The test files are in a different place on master.

The patch is against master, but I wrote the wrong filenames in the
changelog, because I copied part of it from an old commit message.

Helmut