> /home/matsl/work/cedet/cedet_xemacs/semantic/semantic-grammar.el... >>> Error occurred processing semantic-grammar.el: Bad input >>> grammar > I know this is not much to go on but maybe the type of error > tells you something. > > I picked the latest version from CVS to work on. Maybe that was a > mistake? The latest CVS is a fine place to start with CEDET. What you reported is the same problem I ran into. If you use what did compile to start up XEmacs with a typical install (thus loading lots of .el files instead of .elc files) you can compile the grammars using EDE.

A side effect is that I have no idea how to debug this, nor really what the error means. That's where I need some help from an XEmacs expert.

On 09/15/2010 08:21 PM, Stephen J. Turnbull wrote: > Eric M. Ludlam writes: > > > > /home/matsl/work/cedet/cedet_xemacs/semantic/semantic-grammar.el... > > >>> Error occurred processing semantic-grammar.el: Bad > > >>> input grammar

>>>> Eric M Ludlam <eric(a)siege-engine.com&gt; writes:

"Eric M. Ludlam&quot;&lt;eric(a)siege-engine.com&gt; writes: > On 09/16/2010 01:01 AM, Stephen J. Turnbull wrote: >> Eric M. Ludlam writes: >> > On 09/15/2010 08:21 PM, Stephen J. Turnbull wrote: >> > > Eric M. Ludlam writes: >> > > >> > > > > /home/matsl/work/cedet/cedet_xemacs/semantic/semantic-grammar.el... >> > > > >>> Error occurred processing semantic-grammar.el: Bad >> > > > >>> input grammar >> >> BTW, would it be hard to add file/line information to that error >> message, or perhaps quote the line where the error was recognized? > > I provided bad info in my last mail. The grammar file has actually > fully parsed at this point, but it has written nil into the Emacs Lisp > source file. When the data structure created by the parser is > "compiled" it sees the nil and throws this error, so there are no line > numbers at this point. Basically, the whole file is wrong somewhere. > > Thus the error is when the .wy transforms to a .el, some initialization > fails, and the file parses out as empty. Yes. It doesn't even see the terminals in wisent-java-tags.wy. Could you give me a hint where they're actually retrieved? (I keep getting lost in twisty mazes of indirection.) My current thinking is that the work should happen in `semantic-fetch-tags', but the call to `semantic-active-p' there returns nil, so it never does anything.

Thanks for helping with this! "Eric M. Ludlam&quot;&lt;eric(a)siege-engine.com&gt; writes: > The grammar file (.wy) needs to be put into semantic-grammar-mode. > That mode installs parser info into buffer local variables. Those > variables need to be set for semantic-active-p returns true. > > So it could be that the auto-mode-alist is not setup correctly to > install that major-mode on the provided buffer? I've checked: It's set up correctly. > How are you debugging this anyway? With `message' :-) > Things just worked for me when done interactively. Maybe that's the issue: The problem is that, for `semantic-active-p' to return true, `semantic-new-buffer-fcn' needs to run, and it does this via `mode-local': (add-hook 'mode-local-init-hook 'semantic-new-buffer-fcn) `mode-local-init-hook' runs off `mode-local-post-major-mode-change', which in turn runs off `post-command-hook' after `change-major-mode-hook'. ... and `post-command-hook' does not seem to run at this point in the batch build process. Does this make any kind of sense?