[erlang-bugs] compile and syntax_tools disagree

[Ulf Wiger (TN/EAB)]

Maybe it's just I who don't understand how to read the 
doc on abstract forms, but quoted from
http://www.erlang.org/doc/doc-5.5.4/erts-5.5.4/doc/html/part_frame.html

"If C is a function clause ( Ps ) when Gs -> B where Ps 
is a pattern sequence, Gs is a guard sequence and B 
is a body, then 
Rep(C) = {clause,LINE,Rep(Ps),Rep(Gs),Rep(B)}. 

...

A guard Gs is a nonempty sequence of guard tests G_1, ..., 
G_k, and Rep(Gs) = [Rep(G_1), ..., Rep(G_k)]. 

...
If G is true, then Rep(G) = {atom,LINE,true}. 
...
If G is an operator expression E_1 Op E_2, where 
Op is a binary operator, and E_1, E_2 are guard 
expressions, then 
Rep(G) = {op,LINE,Op,Rep(E_1),Rep(E_2)}. "

I read this as 
Rep(C) = {clause,LINE,Rep(Ps),Rep(Gs),Rep(B)}
Rep(Gs) = [Rep(G)|...]
Rep(G) = {op,LINE,Op,Rep(E_1),Rep(E_2)}|...

In other words, as if 
  Guard = [{op,1,'==',{integer,1,1},{var,1,'R'}}]
would actually be valid.

BR,
Ulf W

> -----Original Message-----
> From: Bjorn Gustavsson [mailto:bjorn@REDACTED] 
> Sent: den 3 april 2007 10:31
> To: Ulf Wiger (TN/EAB)
> Cc: erlang-bugs@REDACTED
> Subject: Re: [erlang-bugs] compile and syntax_tools disagree
> 
> As far as I understand, The ERTS User's Guide says that 
> you'll need two layers of lists to represent a guard sequence.
> 
> syntax_tools still accepts the older format of guards (before 
> semicolons in guards were introduced), so that it can still 
> be used for the abstract format in old Beam files.
> 
> So as I see, there is no bug.
> 
> /Björn
> 
> "Ulf Wiger \(TN/EAB\)" <ulf.wiger@REDACTED> writes:
> 
> > I'm getting a bit cross-eyed, but this has got to be a bug 
> *somewhere*:
> > 
> > 12> M = 
> [{attribute,1,module,m},{attribute,1,export,[{foo,1}]}|Fs] ++
> > [{eof,1}]. 
> > [{attribute,1,module,m},
> >  {attribute,1,export,[{foo,1}]},
> >  {function,1,
> >            foo,
> >            1,
> >            [{clause,1,
> >                     [{var,1,'R'}],
> >                     [{op,1,'==',{var,1,'R'},{integer,1,1}}],
> >                     [{atom,1,true}]}]},  {eof,1}]
> > 13> compile:forms(M).
> > .:none: internal error in expand_module; crash reason: 
> > {function_clause,
> >                   [{erl_expand_records,
> >                        guard_tests1,
> >                        [{op,1,'==',{var,1,'R'},{integer,1,1}},
> > ...
> >                    {erl_expand_records,guard_tests,2},
> >                    {erl_expand_records,guard,2},
> >                    {erl_expand_records,clauses,2},
> >                    {erl_expand_records,forms,2},
> >                    {erl_expand_records,forms,2},
> >                    {erl_expand_records,module,2},
> >                    {sys_pre_expand,module,2}]} error
> > 14> M1 = 
> [{attribute,1,module,m},{attribute,1,export,[{foo,1}]}|Fs1] 
> > 14> ++
> > [{eof,1}].
> > [{attribute,1,module,m},
> >  {attribute,1,export,[{foo,1}]},
> >  {function,1,
> >            foo,
> >            1,
> >            [{clause,1,
> >                     [{var,1,'R'}],
> >                     [[{op,1,'==',{var,1,'R'},{integer,1,1}}]],
> >                     [{atom,1,true}]}]},  {eof,1}]
> > 15> compile:forms(M1).
> > {ok,m,
> >  
> > 
> <<70,79,82,49,0,0,1,184,66,69,65,77,65,116,111,109,0,0,0,50,0,0,0,6,1,
> > 10
> > 9,
> >       ...>>}
> > 
> > When pretty-printing the two form lists using 
> erl_prettypr:format/2, 
> > you get the same output, and according to the ERTS User's 
> Guide, the 
> > two guard representations should be equivalent. A list of 
> lists in the 
> > Guard represents a disjunction, and a list of Guard expressions 
> > represents a conjunction. In this case, there's only one guard expr.
> > 
> > 6> erl_prettypr:format(erl_syntax:form_list(Fs)).
> > "foo(R) when R == 1 -> true."      
> > 6> erl_prettypr:format(erl_syntax:form_list(Fs1)).
> > "foo(R) when R == 1 -> true."      
> > 
> > It seems to me as if the bug is in the compiler (version 4.4.3).
> > 
> > Forgive me for not diving into the compiler source code to 
> verify this 
> > claim. Time to put the kids to sleep.
> > 
> > BR,
> > Ulf W
> > _______________________________________________
> > erlang-bugs mailing list
> > erlang-bugs@REDACTED
> > http://www.erlang.org/mailman/listinfo/erlang-bugs
> 
> --
> Björn Gustavsson, Erlang/OTP, Ericsson AB
>