caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
From: "Jocelyn Sérot" <jocelyn.serot@uca.fr>
To: Gabriel Scherer <gabriel.scherer@gmail.com>
Cc: caml users <caml-list@inria.fr>
Subject: Re: [Caml-list] Problem with ocamldebug 4.06.0
Date: Thu, 19 Jul 2018 18:24:21 +0200	[thread overview]
Message-ID: <0B409D6E-DB6F-491F-8D95-67F76700971F@uca.fr> (raw)
In-Reply-To: <CAPFanBGPCY-Uq4d6bRQyzy7==gJDzOGpn3xwk6jpcL9R5oPSFg@mail.gmail.com>

Hi Gabriel, and thanks for your answers.

Le 19 juil. 2018 à 17:57, Gabriel Scherer <gabriel.scherer@gmail.com> a écrit :

> Hi Jocelyn,
> 
> 
>> i) is there something special to do when using ocamldebug in a multi-directory source file context ?
> 
> I don't think so. On the other hand, dynamic linking (with Dynlink) is
> not currently supported, in the sense
> that no debug information will be available for the dynlinked code.
> (It looks like this is not the concern in your case.)
> 
No, no dynamic linking in this case.

>> ii) has someone already encountered this problem
> 
> We recently discussed a bug in ocamldebug related to the interplay
> between debug information and inlining ( MPR#7554 ), for which there
> is a fix by Thomas Refis in the current development version
> (4.08+dev), but not in any released version of OCaml -- not 4.07. The
> bug is general, but in particular it affects the variables foo, bar,
> baz in the following code pattern:
> 
>  let (foo, bar, baz) = (e1, e2, e3) in ...
> 
> If the source program where you observe your problem is publicly
> available, would you mind giving a pointer to the problematic
> variables occurrences?
> 
I’ll check but i’m pretty sure the problem showed for other kinds of pattern bindings.

> I would encourage you to file a problem report on the OCaml bugtracker
> ( https://caml.inria.fr/mantis ) to move forward, especially if you
> have a repro-case or can at least point us on a publicly-available
> version of the source code.

The code is indeed publicly available : https://github.com/jserot/rfsm.
But i only noticed the ocamldebug bug with a development version which i have not pushed yet. 
It may be present with the currently available version though.
I’ll prepare a file problem report with a detailed sequence of instructions to repeat the problem.

> 
>> iii) is there a way to track down where / why the exception was launched (i.e. to debug the debugger)
> 
> In theory you can use OCAMLRUNPARAM=b when running the compiler or
> debugger, and you get a backtrace. That said, the compiler codebase
> was not written with backtraces in mind, and it has a tendency to
> catch and re-raise exceptions in a way that does not preserve
> backtraces -- so the trace you observe may be inaccurate. This is
> something that we could and should improve -- François Bobot proposed
> a change that notably improves the statu quo (GPR#374) but it has not
> been merged upstream yet.

Ah, i haven’t thought of running ocamldebug with OCAMLRUNPARAM set (thought it only applied to programs compiled by the user). Will try.

> 
>> iv) is there a chance that the pb is fixed in 4.07.0
> 
> Probably not (if you say it's not in the Changes), but with your help
> it could very well be fixed in 4.08 :-)

Anyway, switching to 4.08 will do no harm ;)

Thanks again

Jocelyn

> On Thu, Jul 19, 2018 at 4:06 PM Jocelyn Sérot <jocelyn.serot@uca.fr> wrote:
>> 
>> Hello,
>> 
>> I’m experiencing problems when using ocamldebug on a large program.
>> Namely, the debugger quits launching an « Not_found » exception as soon as i try to visualize some program variables (print) or dump thef frame
>> 
>> (ocd) …
>> (ocd) print t
>> Removed breakpoint 1 at 958036: file vhdl.ml, line 113, characters 19-30
>> Uncaught exception: Not_found
>> 
>> OCaml-Debugger exited abnormally with code 2
>> 
>> I’ve started to encounter this problem when i recrafted the program in order to use ocamlbuild and distribute the source code in several directories.
>> What is annoying is that the problem is not systematic it seems to dependi on the variable to be printed (or at least on the module where it is defined) - which makes isolating a minimum triggering program difficult :(
>> 
>> So my questions are
>> 
>> i) is there something special to do when using ocamldebug in a multi-directory source file context ? I of course tried to add « _build » and « subdir/_build » to the debug « path » using the directory command, w/o success
>> 
>> ii) has someone already encountered this problem
>> 
>> iii) is there a way to track down where / why the exception was launched (i.e. to debug the debugger)
>> 
>> iv) is there a chance that the pb is fixed in 4.07.0 (apparently no since nothing is told about ocamldebug in the release doc).
>> 
>> Any help appreciated
>> 
>> Cheers
>> 
>> Jocelyn
>> 


-- 
Caml-list mailing list.  Subscription management and archives:
https://sympa.inria.fr/sympa/arc/caml-list
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
Bug reports: http://caml.inria.fr/bin/caml-bugs

      reply	other threads:[~2018-07-19 16:25 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-07-19 14:05 Jocelyn Sérot
2018-07-19 15:57 ` Gabriel Scherer
2018-07-19 16:24   ` Jocelyn Sérot [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=0B409D6E-DB6F-491F-8D95-67F76700971F@uca.fr \
    --to=jocelyn.serot@uca.fr \
    --cc=caml-list@inria.fr \
    --cc=gabriel.scherer@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).