Re: Latest set of patches to the java-push branch

[Akim Demaille]

Hi Dennis,

Please, always keep the list in CC for the records.

Le 21 juin 2013 à 17:38, Dennis Heimbigner <address@hidden> a écrit :

> As for YYMORE, changing it to something else is fine with me.
> I just thought it was in line with e.g. YYABORT.
> 
> If memory serves. the reason I reset yyerrstatus is because
> it looked to me like the pull loop did it on each pass.
> If doing it as part of the initialize code makes
> those errors go away and does not break anything else,
> then go ahead and change it (as you have done).
> 
> =Dennis Heimbigner
> Unidata
> 
> Akim Demaille wrote:
>> Le 13 juin 2013 à 18:13, Dennis Heimbigner <address@hidden> a écrit :
>>> Actually ignore last meesages. I trust my
>>> test cases enough that if they are passing
>>> then it is ok, so go ahead and push.
>> hi Dennis,
>> I'd like to understand why you reset yyerrstatus
>> in the push-parser wrapper that provide the usual
>> pull interface.  I'm suggesting the following change.
>> Actually, it is requested by the test suite.
>> The test suite is designed to run pull parser tests in
>> push mode with pull emulation.  In other words, any test
>> run for regular pull parsers is run another time, but with
>> a push parser.  This is run via
>>      make maintainer-push-check
>> For instance run:
>> $ make check TESTSUITEFLAGS='-d 388 BISON_USE_PUSH_FOR_PULL=1 -v -x'         
>>                                       and you'll see that there are two 
>> parse error messages instead of
>> only one:
>> --- expout   2013-06-21 11:10:10.000000000 +0200
>> +++ 
>> /Users/akim/src/gnu/bison/_build/48d-debug/tests/testsuite.dir/at-groups/388/stdout
>>       2013-06-21 11:10:10.000000000 +0200
>> @@ -1 +1,2 @@
>> 1.3: syntax error, unexpected '/', expecting number or '-' or '(' or '!'
>> +1.4: syntax error, unexpected number, expecting '\n'
>> 388. java.at:360:  FAILED (java.at:360)
>> Also, why did you name your special value YYMORE?  In C, it is named
>> YYPUSH_MORE.  I can understand why you chose the shorter name, yet
>> I try to keep Bison consistent.
>> commit 1d678854a43d4156e28f47942fcd69adfa04ca96
>> Author: Akim Demaille <address@hidden>
>> Date:   Fri Jun 21 11:17:05 2013 +0200
>>    java: push: do not reset the error counter
>>        * data/lalr1.java (parse): here, when in push-pull is in "both" mode.
>>    This breaks the test suite, for instance
>>    make check TESTSUITEFLAGS='-d 388 BISON_USE_PUSH_FOR_PULL=1'.
>>    More generally make maintainer-push-check.
>> diff --git a/data/lalr1.java b/data/lalr1.java
>> index d226ce3..1598823 100644
>> --- a/data/lalr1.java
>> +++ b/data/lalr1.java
>> @@ -899,7 +899,6 @@ b4_both_if([[
>> ]b4_locations_if([dnl
>>         b4_location_type yyloc = new b4_location_type (yylexer.getStartPos 
>> (),
>>                                               yylexer.getEndPos ());])[
>> -        this.yyerrstatus_ = 0;
>>         ]b4_locations_if([status = push_parse(token,lval,yyloc);],[
>>         status = push_parse(token,lval);])[
>>       } while (status == YYMORE);