CEDET

[...]
> Thanks!  I had started the process, but didn't get very far.
> 
> I also have a patch for the old .bnf file.  It was badly incompatible
> with the .by file I had created, but perhaps not so for yours.  I
> will check it into the .bnf file so you can see the diffs easily.

Thanks! I got the diffs, and applied them to erlang.by.
It was easy ;-)

>>So, we can remove BNF support from Semantic 2.0.
> 
> 
> I know of no other .bnf files not already converted.

Maybe could we remove erlang.bnf and semantic-bnf.el now?

David

Re[2]: [CEDET-devel] A patch for semantic-util-modes.el From: Eric M. Ludlam <eric@si...> - 2003-08-28 15:51 >>> David PONCE <david.ponce@...> seems to think that: [ ... ] >>>So, we can remove BNF support from Semantic 2.0. >> >> >> I know of no other .bnf files not already converted. > >Maybe could we remove erlang.bnf and semantic-bnf.el now? [ ... ] That seems like a good idea. Loading bnf files currently throws an error for me anyway. ;) When I run make w/ erlang.by enabled, I get the following error: ----- choochoo:bovine> make emacs -batch -q -l grammar-make-script -f semantic-grammar-batch-build-packages c.by make.by scheme.by erlang.by Package `semantic-c-by' is up to date. Package `semantic-make-by' is up to date. Package `semantic-scm-by' is up to date. Package `semantic-erlang-by' is up to date. While compiling toplevel forms in file /home/zappo/cedet/semantic/bovine/semantic-erlang-by.el: !! error (("Variable binding depth exceeds max-specpdl-size")) make: *** [by] Error 1 ------ Any ideas? Thanks Eric -- Eric Ludlam: zappo@..., eric@... Home: http://www.ludlam.net Siege: http://www.siege-engine.com Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org

Re: [CEDET-devel] A patch for semantic-util-modes.el From: David Ponce <david.ponce@wa...> - 2003-08-28 20:13 > When I run make w/ erlang.by enabled, I get the following error: > [...] > While compiling toplevel forms in file > /home/zappo/cedet/semantic/bovine/semantic-erlang-by.el: > !! error (("Variable binding depth exceeds max-specpdl-size")) > make: *** [by] Error 1 I got the same error when I tried to byte-compile this file using both latest GNU Emacs 21.3.50 and old GNU Emacs 20.7.1 :-( I also tried to byte compile the previous version of semantic-erlang.el (before conversion to BY format), and got the same error! I eliminated things I finally found that the byte-compiler fails on the parse table expression in `semantic-erlang-by--parse-table' defconst. I just kept that defconst in semantic-erlang-by.el and when I set the max-lisp-eval-depth to 1000, I got these compiler warnings: Compiling file c:/TEMP/semantic-erlang-by.el at Thu Aug 28 22:12:07 2003 semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' with too few arguments semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' with too few arguments semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' with too few arguments semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' with too few arguments semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' with too few arguments semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' with too few arguments Weird. I don't understand what happens here. Maybe a byte-compiler bug? I continue to investigate... David

Re[2]: [CEDET-devel] A patch for semantic-util-modes.el From: Eric M. Ludlam <eric@si...> - 2003-08-29 03:34 >>> David Ponce <david.ponce@...> seems to think that: > > When I run make w/ erlang.by enabled, I get the following error: > > >[...] > > While compiling toplevel forms in file > > /home/zappo/cedet/semantic/bovine/semantic-erlang-by.el: > > !! error (("Variable binding depth exceeds max-specpdl-size")) > > make: *** [by] Error 1 > >I got the same error when I tried to byte-compile this file using both >latest GNU Emacs 21.3.50 and old GNU Emacs 20.7.1 :-( > >I also tried to byte compile the previous version of >semantic-erlang.el (before conversion to BY format), and got the same >error! > >I eliminated things I finally found that the byte-compiler fails on >the parse table expression in `semantic-erlang-by--parse-table' defconst. > >I just kept that defconst in semantic-erlang-by.el and when I set the >max-lisp-eval-depth to 1000, I got these compiler warnings: > >Compiling file c:/TEMP/semantic-erlang-by.el at Thu Aug 28 22:12:07 2003 >semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments >semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments >semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments >semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments >semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments >semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' > with too few arguments > >Weird. I don't understand what happens here. Maybe a byte-compiler >bug? I continue to investigate... [ ... ] Fascinating. This is a common warning in many of our files now. I first noticed when I started changing names and making old function names obsolete. Perhaps it is somehow related to that. Eric -- Eric Ludlam: zappo@..., eric@... Home: http://www.ludlam.net Siege: http://www.siege-engine.com Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org

Re: [CEDET-devel] A patch for semantic-util-modes.el From: David PONCE <david.ponce@wa...> - 2003-08-29 07:59 Hi Eric, [...] >>I just kept that defconst in semantic-erlang-by.el and when I set the >>max-lisp-eval-depth to 1000, I got these compiler warnings: >> >>Compiling file c:/TEMP/semantic-erlang-by.el at Thu Aug 28 22:12:07 2003 >>semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments >>semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments >>semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments >>semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments >>semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments >>semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' >> with too few arguments [...] > Fascinating. This is a common warning in many of our files now. I > first noticed when I started changing names and making old function > names obsolete. Perhaps it is somehow related to that. I found what caused the above warnings: it was a missing parameter in calls to `semantic-tag-new-type' and `semantic-tag-new-include'. As these functions are inlined, the compiler issued those warnings. I fixed erlang.by and re-generated semantic-erlang-by.el, and the warnings are gone :-) About the "Variable binding depth exceeds max-specpdl-size" error, I think it is caused by the complexity of the grammar table expression. So I updated `semantic-grammar-batch-build-one-package' to give Emacs a little more resources, and that fixed the error. I also fixed a nasty bug I encountered in `semantic-grammar-create-package', when the buffer name of the generated lisp file is "uniquified" by the uniquify library. Now `semantic-grammar-create-package' no more depend on buffer name. Here is the change log: 2003-08-29 David Ponce <david@...> * semantic/semantic-grammar.el (semantic-grammar-buffer-file): New function. (semantic-grammar-package, semantic-grammar-keywordtable) (semantic-grammar-tokentable, semantic-grammar-parsetable) (semantic-grammar-setupfunction, semantic-grammar-header) (semantic-grammar-footer): Use it. (semantic-grammar-batch-build-one-package): Increase some Emacs limits to byte-compile complex grammar tables. * semantic/bovine/erlang.by (record-def): Fix call to TYPE-TAG. (include-dir, include-lib-dir): Fix call to INCLUDE-TAG. * semantic/bovine/semantic-erlang-by.el Re-generated. David

Re[2]: [CEDET-devel] A patch for semantic-util-modes.el From: Eric M. Ludlam <eric@si...> - 2003-08-29 12:21 >>> David PONCE <david.ponce@...> seems to think that: >Hi Eric, > >[...] >>>I just kept that defconst in semantic-erlang-by.el and when I set the >>>max-lisp-eval-depth to 1000, I got these compiler warnings: >>> >>>Compiling file c:/TEMP/semantic-erlang-by.el at Thu Aug 28 22:12:07 2003 >>>semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >>>semantic-erlang-by.el:653:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >>>semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >>>semantic-erlang-by.el:1524:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >>>semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >>>semantic-erlang-by.el:1564:7:Warning: attempt to open-code `anonymous lambda' >>> with too few arguments >[...] >> Fascinating. This is a common warning in many of our files now. I >> first noticed when I started changing names and making old function >> names obsolete. Perhaps it is somehow related to that. > >I found what caused the above warnings: it was a missing parameter in >calls to `semantic-tag-new-type' and `semantic-tag-new-include'. As >these functions are inlined, the compiler issued those warnings. That's great news. Now I'll know how to fix the other warnings that occur all over the place. Is there a way to get a better error I wonder. That message made no sense to me. >I fixed erlang.by and re-generated semantic-erlang-by.el, and the >warnings are gone :-) > >About the "Variable binding depth exceeds max-specpdl-size" error, I >think it is caused by the complexity of the grammar table expression. >So I updated `semantic-grammar-batch-build-one-package' to give Emacs >a little more resources, and that fixed the error. > >I also fixed a nasty bug I encountered in >`semantic-grammar-create-package', when the buffer name of the >generated lisp file is "uniquified" by the uniquify library. Now >`semantic-grammar-create-package' no more depend on buffer name. That all sounds like good stuff. Thanks! Eric -- Eric Ludlam: zappo@..., eric@... Home: http://www.ludlam.net Siege: http://www.siege-engine.com Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org

Re: [CEDET-devel] A patch for semantic-util-modes.el From: David PONCE <david.ponce@wa...> - 2003-08-29 13:23 [...] > That's great news. Now I'll know how to fix the other warnings that > occur all over the place. Is there a way to get a better error I > wonder. That message made no sense to me. [...] Maybe a good solution would be change the tag macro expanders definition from something like: (defun bovine-grammar-INCLUDE-TAG (&rest args) "Expand call to INCLUDE-TAG grammar macro. Return the expanded form. ARGS are the arguments passed to the macro." `(semantic-tag-new-include ,@args)) to something like: (defun bovine-grammar-INCLUDE-TAG (name system-flag &rest attributes) "Expand call to INCLUDE-TAG grammar macro. Return the expanded form that creates a semantic tag of class include. NAME is the name of this include. SYSTEM-FLAG represents that we were able to identify this include as belonging to the system, as opposed to belonging to the local project. ATTRIBUTES is a list of additional attributes belonging to this tag." `(semantic-tag-new-include ,name ,system-flag ,@attributes)) Thus, calling the macro with a wrong number of arguments will fail at lisp generation time. What do you think? David

Re[2]: [CEDET-devel] A patch for semantic-util-modes.el From: Eric M. Ludlam <eric@si...> - 2003-08-29 14:37 >>> David PONCE <david.ponce@...> seems to think that: >[...] >> That's great news. Now I'll know how to fix the other warnings that >> occur all over the place. Is there a way to get a better error I >> wonder. That message made no sense to me. >[...] > >Maybe a good solution would be change the tag macro expanders >definition from something like: > >(defun bovine-grammar-INCLUDE-TAG (&rest args) > "Expand call to INCLUDE-TAG grammar macro. >Return the expanded form. >ARGS are the arguments passed to the macro." > `(semantic-tag-new-include ,@args)) > >to something like: > >(defun bovine-grammar-INCLUDE-TAG (name system-flag &rest attributes) > "Expand call to INCLUDE-TAG grammar macro. >Return the expanded form that creates a semantic tag of class include. >NAME is the name of this include. >SYSTEM-FLAG represents that we were able to identify this include as belonging >to the system, as opposed to belonging to the local project. >ATTRIBUTES is a list of additional attributes belonging to this tag." > `(semantic-tag-new-include ,name ,system-flag ,@attributes)) > >Thus, calling the macro with a wrong number of arguments will fail at >lisp generation time. [ ... ] That's a good idea. I've fixed all the avoidable compiler warnings for the semantic lisp code not in bovine or wisent. Now I just need time to check it all in. Eric -- Eric Ludlam: zappo@..., eric@... Home: http://www.ludlam.net Siege: http://www.siege-engine.com Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org

[CEDET-devel] Use of explicit arglist in grammar macro expanders From: David Ponce <david.ponce@wa...> - 2003-08-31 09:19 Hi Eric, [...] >>Maybe a good solution would be change the tag macro expanders >>definition from something like: >> >>(defun bovine-grammar-INCLUDE-TAG (&rest args) [...] >>to something like: >> >>(defun bovine-grammar-INCLUDE-TAG (name system-flag &rest attributes) [...] > > That's a good idea. I rewrote the bovine and wisent grammar macro expanders to use explicit arguments, and that revealed an inconsistency between the use of the EXPAND macro in scheme.by and erlang.by, and the way it is expanded (proof of the advantage of this new design ;-) It appeared that that inconsistency is there since the old BNF converter. An example from old scheme.bnf: scheme : semantic-list (EXPAND $1 scheme-list 4) ; The EXPAND macro uses an explicit DEPTH argument '4'. The BNF converter expanded that to: `((bovine-toplevel ( semantic-list ,(lambda (vals start end) (semantic-bovinate-from-nonterminal (car (nth 0 vals)) (cdr (nth 0 vals)) 'scheme-list) )) ) ; end scheme Silently ignoring the DEPTH argument! Until now, the BY converter did the same thing. The old erlang.bnf and new erlang.by grammars contain such use of an explicit DEPTH argument, silently ignored by the expander. With the new expander which use explicit arguments: (defun bovine-grammar-EXPAND (bounds nonterm) ...) The lisp generation fails with a "Wrong number of arguments" error. So my question: should I include an new optional argument DEPTH in EXPAND expanders, or should I fix the grammars, because the DEPTH argument wasn't never used? I would prefer to fix the grammars, as they work using the default DEPTH equal to '1'. Also that would be consistent with the EXPANDFULL expander that uses DEPTH=1 by default. What do you think? David

[CEDET-devel] Re[1]: Use of explicit arglist in grammar macro expanders From: Eric M. Ludlam <eric@si...> - 2003-08-31 12:36 >>> David Ponce <david.ponce@...> seems to think that: >Hi Eric, > >[...] > >>Maybe a good solution would be change the tag macro expanders > >>definition from something like: > >> > >>(defun bovine-grammar-INCLUDE-TAG (&rest args) >[...] > >>to something like: > >> > >>(defun bovine-grammar-INCLUDE-TAG (name system-flag &rest attributes) >[...] > > > > That's a good idea. > >I rewrote the bovine and wisent grammar macro expanders to use >explicit arguments, and that revealed an inconsistency between the >use of the EXPAND macro in scheme.by and erlang.by, and the way it is >expanded (proof of the advantage of this new design ;-) > >It appeared that that inconsistency is there since the old BNF >converter. > >An example from old scheme.bnf: > >scheme : semantic-list > (EXPAND $1 scheme-list 4) > ; > >The EXPAND macro uses an explicit DEPTH argument '4'. The BNF >converter expanded that to: > >`((bovine-toplevel > ( semantic-list > ,(lambda (vals start end) > > (semantic-bovinate-from-nonterminal (car (nth 0 vals)) (cdr (nth 0 vals)) 'scheme-list) > )) > ) ; end scheme > >Silently ignoring the DEPTH argument! Until now, the BY converter did the same >thing. > >The old erlang.bnf and new erlang.by grammars contain such use of an >explicit DEPTH argument, silently ignored by the expander. > >With the new expander which use explicit arguments: > >(defun bovine-grammar-EXPAND (bounds nonterm) > ...) > >The lisp generation fails with a "Wrong number of arguments" error. > >So my question: should I include an new optional argument DEPTH in >EXPAND expanders, or should I fix the grammars, because the DEPTH >argument wasn't never used? > >I would prefer to fix the grammars, as they work using the default >DEPTH equal to '1'. Also that would be consistent with the EXPANDFULL >expander that uses DEPTH=1 by default. [ ... ] Actually, the scheme parser misses some forms and I couldn't figure out why. Perhaps the depth inconsistency is why. An inspection indicates that is unlikely however. Removing the argument does make sense if it will simplify the process. I'm glad the new mechanism exposes bugs so readily! Eric -- Eric Ludlam: zappo@..., eric@... Home: http://www.ludlam.net Siege: http://www.siege-engine.com Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org

Re: [CEDET-devel] Re[1]: Use of explicit arglist in grammar macro expanders From: David Ponce <david.ponce@wa...> - 2003-08-31 15:12 > Actually, the scheme parser misses some forms and I couldn't figure > out why. Perhaps the depth inconsistency is why. An inspection > indicates that is unlikely however. Removing the argument does make > sense if it will simplify the process. I did that and checked-in the changes. > I'm glad the new mechanism exposes bugs so readily! Yes it nice to detect bugs as soon as possible ;-) David

1 message has been excluded from this view by a project administrator.