[Vala] Fwd: Name Mangling

[pancake <pancake youterm com>]

Sent from my iPod

Begin forwarded message:

From: James Cox <jamescox84 googlemail com>
Date: July 24, 2009 11:53:27 AM GMT+02:00
To: pancake <pancake youterm com>
Subject: Re: [Vala] Name Mangling

Maybe the only solution (all be it not necessarily a clean one) is to
keep the current behavior, only have the Vala compiler issue an error
that explains what the problem is.

test.vala:#.#-#.#: error: The symbol name `asm' is a C reserved word.
test.vala:#.#-#.#: error: Name mangling error
`my_name_space_myclass_new' has already been used for
`MyNameSpace.MyClass.ctor'
test.vala:#.#-#.#: error: Name mangling error `MyNameSpace.my_class'
has already been used for `MyNameSpace.MyClass'

Something like this would be great obviously the user could then look
into the specific details of the error, and have a better chance of
understanding what's going on.  Rather than trying to figure out what
the C error messages mean.

On Thu, Jul 23, 2009 at 10:46 PM, pancake<pancake youterm com> wrote:

The same happens when using C keywords like 'asm'. I can't imagine a clean

solution for this :/

int asm=0;

Sent from my iPod

On Jul 23, 2009, at 1:00 PM, James Cox <jamescox84 googlemail com> wrote:

Hi Everyone,

Has this been discussed before, or even fixed.  It's to do with the

way Vala handles the mangling of the names of various symbols, and how

these rule can create identical names in the C source code.  I propose

that the Vala compile issues a informative error when such conflicts

arise.

I've been using Vala for a while now.  I've got to say it's absolutely

an amazing idea, to use the C+GObject type system as the runtime

environment for a language is ingenious!  One thing I noticed though

(I appreciate this would be down to some bad design on the

programmer's (user'ss) part) is the way Vala mangles the names of

classes, methods.  It is possible to compile code to the intermediate

C and then have that code fail when the code is compiled by cc due to

redefinition of symbols, that had distinct names in the Vala source

file, but now have the same name in the C source files.  Here are some

examples:

namespace MyNameSpace

{

  class MyClass

  {

  }

  // Matches Mangleing of MyNameSpace.MyClass

  class my_class

  {

  }

}

// Matches Mangleing of MyNameSpace.MyClass

class MyNameSpaceMyClass

{

}

// Redefines the MyNameSpace.MyClass constructor due to matchingh magling.

public void my_name_space_myclass_new()

{

}

I'm not sure of the best way to solve the issues.  Maybe not

guaranteeing the name of a function that is  declared out side of a

class/namespace, an error when mangling matches another symbol.  Maybe

enforce a naming convention for class names, much in the same way as

Ruby...  Personally I think that keeping the same rules for name

mangling is best (keeps API looking like a normal GObject API), but

just make Vala issue a compile time error explaining that a

redefinition has be made and point the programmer to the two (or more)

location where it happened.  I'm not sure about where an when these

problems might arise in normal use, but I think the behavior of the

language should be specified for when these problems occur.

_______________________________________________

Vala-list mailing list

Vala-list gnome org

http://mail.gnome.org/mailman/listinfo/vala-list