[openbeosnetteam] Re: Moving stuff: half done.
- From: Ingo Weinhold <bonefish@xxxxxxxxxxxxxxx>
- To: openbeosnetteam@xxxxxxxxxxxxx
- Date: Wed, 28 Aug 2002 15:39:39 +0200 (MET DST)
On Wed, 28 Aug 2002, Marcus Overhagen wrote:
> This should read:
> > this would be a bad idea. Instead, a much better idea is to include
> > SupportDefs.h
> > at the places where we use sys/types.h
> ..if we need to use types like int32.
>
> > If possible, I'm going to replace the current sys/types.h with the BeOS R5
> > sys/types.h
> > This will break much, but I think this needs to be done and I'm going to
> > fix it then.
> I tried it, I think it is the way to go.
>
> It will also require to at least add
> headers/posix/be_setup.h
> headers/posix/null.h
> headers/posix/size_t.h
> headers/posix/time.h
> and:
> headers/posix/sys/select.h
> headers/posix/sys/sockio.h
>
> Philippe did you forget to add the last two, or am I doing something wrong?
>
> I still think:
> > However I have still a bad feeling about our header problems. We really
> > should
> > work on getting a compilation that is independent from headers in
> > /boot/develop/headers
>
> Is there a way to ignore /boot/develop/headers (other then by renaming it)?
Yes, either the gcc option -nostdinc or the to include all OBOS headers,
which will shadow the ones in /boot/develop/headers. Or both.
> I also thing there is another problem with the current build system.
> In BeOS, you could include any header without worrying where is was,
> but now we have to do things like UsePublicHeaders support;
>
> Can't we make *all public* headers available by default without extra rules?
>
> something like
> current/headers/os current/headers/posix current/headers/os/app
> current/headers/os/device
> current/headers/os/drivers current/headers/os/game ...
There is the rule SetupIncludes, which does exactly this. It was
originally called from all "higher level" rules (e.g. Addon, Server,
SharedLibrary), but I had to comment it out, as some of our headers are
currently incompatible with R5's (e.g. <OS.h> or <sys/types.h>).
> I don't like errors like this one, just because current/headers/os isn't
> included,
> and adding a rule for this to every Jamfile doesn't seem to be an option too.
Yep, but until all of our headers are compatible with the original ones,
we can't do that.
> BTW, is there a build team list, and can somebody please forward it?
There is a build team list, but as Ryan seems to be a bit low on spare
time, it wouldn't help that much to take it to that list. And for I'm
lurking here anyway, feel free to just post requests on this list.
> PS: don't fear, I'm not going to commit anything that will break the build of
> the kernel.
You would be the one that would be smacked. ;-)
CU, Ingo
Other related posts:
