On Mon, Apr 09, 2007 at 11:58:03AM +0200, Stefano Zacchiroli wrote:
> [ Ob: debian-ocaml-maint, look at the end of this mail ]
> 
> tags 415194 + wontfix
> thanks
> 
> On Fri, Mar 16, 2007 at 06:55:27PM -0400, Ivan Jager wrote:
> > The bytecode files are currently compiled without debugging support.
> > This makes it hard to debug other code that might be called by it. Eg, a
> > function passed to List.iter.
> 
> In the Debian folklore of libraries (at large, not OCaml-specific)
> that's a feature, not a bug. Indeed usually libraries do not contain
> debugging symbols and where is deemed appropriate an extra -dbg library
> package is built containing debugging information.
> 
> > Since the standard libraries are compiled with debugging support it
> > seems like it would make sense to do the same for others. Yes, it might
> > be slightly slower, but anyone who cares about speed will probably be
> > compiling native code anyways.
> 
> In the specific case of OCaml libraries I think we never discussed the
> issue and therefore I think the standard libraries just happen to be
> compiled that way (probably because they are compiled that way upstream)
> without any particular reason.
> 
> So, at the moment I'm tagging this bug report as wontfix, but diverting
> the more general question of "should we mandate inclusion of debugging
> symbols in OCaml bytecode libraries"? to the debian-ocaml-maint mailing
> list. On one hand we should be consistent with the general library
> philosophy of not including debugging symbols in packages other than
> -dbg. On the other it is true that a user willing to have performances
> will use native code libraries, but is still true that native code
> libraries are not available everywhere ...

One interesting question here, is what is the cost of adding those debugging
symbols ? 

Is this cost a performance hit, or only a size increase ? 

Is anyone familiar with how debugging is implemented in ocaml ?

Friendly,

Sven Luther


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to