This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [treelang] Make treelang work with --enable-mapped-location

Zack Weinberg <> writes:

> This is hard to review due to the large number of unrelated changes.
> I will preapprove one patch which fixes comment formatting and does
> nothing else, and another patch which converts abort() to gcc_assert()
> or gcc_unreachable() as appropriate and does nothing else.  [Yes, I
> want that as two separate patches.]  Once you've done that, please
> resubmit these changes.

 Just to be sure, you want:
  1) A patch to fix comment formatting [ preapproved ]
  2) A patch to convert to gcc_assert and gcc_unreachable [ preapproved ]
  3) A patch to change to error/warning and mapped locations

> Also, some nitpicks:
> 1) The closing */ of a comment goes at the end of the last line of
>    comment text, not on a line by itself.  I'll grant an exception
>    for cases where the last thing in the comment is example code or an
>    ASCII-art diagram, so it would look awful if you did that.
> 2) In error() and related functions, %<%X%> can be written %qX.  I
>    believe this works for every formatting escape %X supported by
>    error(), but I could be wrong (Joseph Myers or Gabriel Dos Reis
>    would know for certain).

 Humm, I didn't know this.  I learn something everyday.  I'm using .*s,
because the length of the token's string is specified.  I didn't know if this
is an optimization or needed yet.  It seems what is currently done is a the
lexem is turned into a tree, then the IDENTIFER_POINTER is put into the token
struct.  That seems overly complicated to me.
> >  PING:
> This one is OK as-is.
> zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]