Re: why this is shift/reduce?

[Carl Cerecke]

Jibin Han wrote:
> Hello,
>     Bison complains the following shift/reduce error at
> chip_param_define section(the last part), I am confused here, could you
> tell me why it is?
>
> board:chips {}
>
> chips: chip {}
>   |
>   chips chip {}
>
> chip: BEGIN_FPGA chip_defines cores END {}
>
> chip_defines: chip_param_define {}
>   |
>   chip_defines chip_param_define {}
>
> chip_param_define: /* empty */
>   |
>   PARAMETER_INSTANCE  CHIP_INSTANCE_NAME {}
>   |
>   DEVICE CHIP_DEVICE_NAME {}


The problem is the "chip_param_define -> /* empty */" rule
in combination with the fact that a chip_param_define
can follow another chip_param_define (from the definition
of chip_define)


The generated parser, when it sees, for example:
PARAMETER_INSTANCE  CHIP_INSTANCE_NAME
could either shift PARAMETER_INSTANCE, or reduce
using the "chip_param_define -> /* empty */" rule,
and then shift PARAMETER_INSTANCE. In fact, you
could reduce "chip_param_define -> /* empty */"
an abitrary number of times before you decided to
shift.

Confused? Draw out the derivation tree twice on paper.
Once without using the "chip_param_define -> /* empty */"
rule, and once using it.

The solution I would recommend in this case is
to change the chip_defines rule to derive 0 or more
chip_param_defines (rather than 1 or more like it is
now), and change chip_param_define so it cannot
derive /* empty */.

Cheers,
Carl.