Version française
Home     About     Download     Resources     Contact us    
Browse thread
weird behavior with camlp4o
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: forum@x...
Subject: Re: [Caml-list] weird behavior with camlp4o
Selon Nicolas Pouillard <nicolas.pouillard@gmail.com>:

> Excerpts from xclerc's message of Wed Apr 09 09:48:22 +0200 2008:
> > Selon Olivier Andrieu <oandrieu@gmail.com>:
> >
> > > On Wed, Apr 9, 2008 at 9:29 AM,  <forum@x9c.fr> wrote:
> > > > Selon "Andrew I. Schein" <andrew@andrewschein.com>:
> > > >  > Greetings list -
> > > >  >
> > > >  > I was playing around with OCaml 3.10.2 camlp4o like this:
> > > >  >
> > > >  > camlp4o pa_breakcont.cmo sample1.ml
> > > >  >
> > > >  > with my macro pa_breakcont.cmo and got the expected macro
> translation
> > > >  > printed to my terminal.  However, when I type:
> > > >  >
> > > >  > camlp4o pa_breakcont.cmo sample1.ml > out.ml
> > > >  >
> > > >  > out.ml contains binary output.  Am I misusing camlp4o?
> > > >
> > > >  I have encountered the same problem a few days ago while working on
> > > >  Ocaml-Java to make it camlp4-compatible.
> > > >
> > > >  The fact is that the kind of output (binary dump of abstract tree or
> > > >  source code in textual form) is chosen according to the nature of the
> > > >  output file descriptor. If the output file descriptor denotes a tty
> > > >  then the textual form is chosen, otherwise the binary form is chosen.
> > > >
> > > >  That being said, I don't know what is the rationale of this choice,
> > > >  as I have not come up with a use case for the binary form.
> > >
> > > It's simply more efficient for ocamlc or ocamlopt when camlp4 is
> > > called via the -pp option: no need to pretty-print and then reparse
> > > the source.
> >
> > Well, this is what I thought at first but, if I am not mistaken, when
> > you use the '-pp' option of a compiler, a command of the following form
> > is executed: "camlp4XXX source-file.ml > tmp-file.ml".
> > Then the "tmp-file.ml" is actually compiled instead of the
> "source-file.ml".
> >
> > So, it seems that compilers go through the textual form.
>
> Nop, the compiler knows both binary and textual formats.

My mistake. Sorry for the noise.
In fact, I mixed up the actual behaviour of camlp4 with the awful
hack I set up to make OCaml-Java camlp4-aware. As I don't know any
pure-Java way to know if a file descriptor denotes a tty, I just
pretend that descriptor "1" is always a tty. As a consequence, camlp4
(when run by OCaml-Java) always outputs the textual form. This in turn
implies that compilers using the '-pp' option will go through the
textual form.

This seemed to work like a charm but I totally missed the location problem
you point out. This should make me craft another hack ...
Thanks for the information.


Xavier Clerc

PS: the 'camlp4' patches for OCaml-Java are not currently applied to the
released version
PPS: if anyone knows a reliable pure-Java way to determine if a Java output
stream is related to a tty, I would just avoid a terrible hack ...