Re: [Vala] bugs

[Jürg Billeter <j bitron ch>]

On Sat, 2007-09-08 at 20:45 +0200, pancake wrote:
> BTW I have noticed the -O flag in valac, that it's just the same as passing
> it to gcc. The problem I see here is that you're forced to place a space
> between the flag and the argument. Hardcoding some flags like -O0 -O1 -O3 ..
> would be more closer to gcc.
>
> Talking about design patterns. This is a bad idea because you are dupping
> LOCs without needing to. There the -X flag to directly pass flags to gcc,
> so -O is not necessary there. So you can:
>
>  $ valac -o foo foo.vala -X -O3
>
> which is the same as:
>
>  $ valac -o foo foo.vala -O 3
>
> For me the first way looks cleaner and more orthogonal :)

Yes, you're probably right. I've added the -O option before the -X
option, IIRC, that's why it's like that now.

> The same can be applied to the -g flag

Not quite, the -g tells valac to emit #line directives, for example.

> The only way I can understand the existence of these flags is to keep operation
> consistence between different compilers (gcc, tcc, msvc, ...) wrapping these
> flags to the correct syntax for the target compiler.
>
> BTW the best way to do this (imho) is to create another program that acts
> as a meta-compiler wrapper for gcc/icc/.. and make valac call the meta-compiler
> and the meta-compiler call the desired compiler with the right flags and environment
> variables.

I don't want to make half a build-system out of valac, the -X option
should work fine the same way it does for C compilation.

> A little question... is there any way to browse namespaces classes and methods?
> a cmdline tool to query this will be nice for an IDE (or vim ;) integration for
> method autocompletion and so.

You can do that via libvala, although it's probably too slow for IDE
integration right now as it has to parse the VAPI files every time -
unless you just keep the symbol tree in memory. A binary - possibly
mmap-able - cache file might make sense here.

Jürg