RBuild woes...

[Alex Ionescu]

Hi,

Here is my beef with rbuild. I can't fix most of these as I don't have 
the skills/knowledge of rbuild.

1) Project link flags are applied at the end of the command line...this 
makes -Wl optimization flags not work.
2) DBG = 0 and GCC 4 creates a load of new warnings that weren't there 
before. For example, I can't build ntoskrnl anymore. This is because 
it's detecting "uninitialized variables" which are actually being 
initalized by the macros themselves (see mm/pe.c and elf.inc.h). This 
did not happen before rbuild.
3) Strip isn't being executed for dbg = 0 anymore... I can add it but I 
have no idea how to make it dbg = 0 only.

Not sure:
4) Is the OPTIMIZED envirovar still taken into account? I didn't seem to 
see so.

Best regards,
Alex Ionescu

[Casper Hornstrup]

>   -----Original Message-----
>  From: ros-dev-bounces(a)reactos.com [mailto:[e-mail]] On Behalf Of Alex
> Ionescu
>  Sent: 28. maj 2005 20:26
>  To: ReactOS Development List
>  Subject: [ros-dev] RBuild woes...
>  
>  Hi,
>  
>  Here is my beef with rbuild. I can't fix most of these as I don't have
>  the skills/knowledge of rbuild.
>  
>  1) Project link flags are applied at the end of the command line...this
>  makes -Wl optimization flags not work. 
Will fix.

>   2) DBG = 0 and GCC 4 creates a load of new warnings
> that weren't there
>  before. For example, I can't build ntoskrnl anymore. This is because
>  it's detecting "uninitialized variables" which are actually being
>  initalized by the macros themselves (see mm/pe.c and elf.inc.h). This
>  did not happen before rbuild. 
I have no idea what is different. Is the gcc command line different?

>   3) Strip isn't being executed for dbg = 0
> anymore... I can add it but I
>  have no idea how to make it dbg = 0 only. 
Will fix.

>   
>  Not sure:
>  4) Is the OPTIMIZED envirovar still taken into account? I didn't seem to
>  see so. 
I see no need for the OPTIMIZED configuration. To me it is only another
source of build errors.

Casper

[Alex Ionescu]

Casper Hornstrup wrote:

>  
>   
>
> >  Not sure:
> > 4) Is the OPTIMIZED envirovar still taken into account? I didn't seem to
> > see so.
> >     
> >  
> I see no need for the OPTIMIZED configuration. To me it is only another
> source of build errors.
>
> Casper
>
>   
>  
I'm sure Thomas will come screaming when he sees this :)

I don't use it myself, but because of gcc's stupidity, it does not 
detect uninitialized variables unless you optimize. What this means is 
that we're all merrily working on DBG = 1, and when it's release time, 
optimizations are turned on and everything breaks because uninitialized 
variables are detected. Plus, it's actually a good thing to find them, 
because it makes finding bugs easier. So OPTIMIZED was created to allow 
-Os to be used even if it's a debug build, so that such errors can be 
detected.

So it's actually a source of eliminiating build errors. But, personally, 
I don't use it, and only Thomas begged me for it...

Best regards,
Alex Ionescu

[Casper Hornstrup]

>   -----Original Message-----
>  From: ros-dev-bounces(a)reactos.com [mailto:[e-mail]] On Behalf Of Alex
> Ionescu
>  Sent: 28. maj 2005 22:25
>  To: ReactOS Development List
>  Subject: Re: [ros-dev] RBuild woes...
>  
>  Casper Hornstrup wrote:
>  
>  I'm sure Thomas will come screaming when he sees this :)
>  
>  I don't use it myself, but because of gcc's stupidity, it does not
>  detect uninitialized variables unless you optimize. What this means is
>  that we're all merrily working on DBG = 1, and when it's release time,
>  optimizations are turned on and everything breaks because uninitialized
>  variables are detected. Plus, it's actually a good thing to find them,
>  because it makes finding bugs easier. So OPTIMIZED was created to allow
>  -Os to be used even if it's a debug build, so that such errors can be
>  detected.
>  
>  So it's actually a source of eliminiating build errors. But, personally,
>  I don't use it, and only Thomas begged me for it...
>  
>  Best regards,
>  Alex Ionescu 
So always build with optimization.

Casper

[Thomas Weidenmueller]

Casper Hornstrup wrote:
>   So always build with optimization. 
That's what I usually do, i always build with optimizations, even for
DBG:=1. The reasons were already mentioned, so unless optimizations are
enabled by default (those with slower machines might oppose that) there
should be at least an option to enforce optimizations.

Best Regards,
Thomas

[Casper Hornstrup]

>   -----Original Message-----
>  From: ros-dev-bounces(a)reactos.com [mailto:[e-mail]] On Behalf Of
> Thomas
>  Weidenmueller
>  Sent: 28. maj 2005 22:43
>  To: ReactOS Development List
>  Subject: Re: [ros-dev] RBuild woes...
>  
>  Casper Hornstrup wrote:
> >   So always build with optimization. 
>  
>  That's what I usually do, i always build with optimizations, even for
>  DBG:=1. The reasons were already mentioned, so unless optimizations are
>  enabled by default (those with slower machines might oppose that) there
>  should be at least an option to enforce optimizations.
>  
>  Best Regards,
>  Thomas 
So people with slow machines are allowed to break building? ;-)

Casper

[Alex Ionescu]

Casper Hornstrup wrote:

>  
>   
>
> >  2) DBG = 0 and GCC 4 creates a load of new
> > warnings that weren't there
> > before. For example, I can't build ntoskrnl anymore. This is because
> > it's detecting "uninitialized variables" which are actually being
> > initalized by the macros themselves (see mm/pe.c and elf.inc.h). This
> > did not happen before rbuild.
> >     
> >  
> I have no idea what is different. Is the gcc command line different?
>
>   
>  
I'll revert and check. I think it's related to include files or 
something. I'll check the cmd line.

Best regards,
Alex Ionescu

[Casper Hornstrup]

>   -----Original Message-----
>  From: ros-dev-bounces(a)reactos.com [mailto:[e-mail]] On Behalf Of Alex
> Ionescu
>  Sent: 28. maj 2005 20:26
>  To: ReactOS Development List
>  Subject: [ros-dev] RBuild woes...
>  
>  Hi,
>  
>  Here is my beef with rbuild. I can't fix most of these as I don't have
>  the skills/knowledge of rbuild.
>  
>  1) Project link flags are applied at the end of the command line...this
>  makes -Wl optimization flags not work. 
That is bad because -lgcc need to be last otherwise you get linker errors.

Casper

[Alex Ionescu]

Casper Hornstrup wrote:

> >  
> > Hi,
> >
> > Here is my beef with rbuild. I can't fix most of these as I don't have
> > the skills/knowledge of rbuild.
> >
> > 1) Project link flags are applied at the end of the command line...this
> > makes -Wl optimization flags not work.
> >     
> >  
> That is bad because -lgcc need to be last otherwise you get linker errors.
>
> Casper
>
>   
>  
Hi,

I don't understand what you mean :(. I am saying that Project link flags 
should not be applied at the end.... or are you saying lgcc is a project 
link flag, and if you make link flags at the beginning then lgcc won't 
be first anymore?

In that case then a "Secondary" project link flag setting must be 
made...or something... I don't know :/

Thanks for helping btw.

Best regards,
Alex Ionescu

[Casper Hornstrup]

>   -----Original Message-----
>  From: ros-dev-bounces(a)reactos.com [mailto:[e-mail]] On Behalf Of Alex
> Ionescu
>  Sent: 28. maj 2005 20:26
>  To: ReactOS Development List
>  Subject: [ros-dev] RBuild woes...
>  
>  Hi,
>  
>  Here is my beef with rbuild. I can't fix most of these as I don't have
>  the skills/knowledge of rbuild.
>  
>  3) Strip isn't being executed for dbg = 0 anymore... I can add it but I
>  have no idea how to make it dbg = 0 only. 
The image is stripped AFAICS.

Casper

[Alex Ionescu]

Alex Ionescu wrote:

>   Hi,
>
>  Here is my beef with rbuild. I can't fix most of these as I don't have 
>  the skills/knowledge of rbuild.
>
>  1) Project link flags are applied at the end of the command 
>  line...this makes -Wl optimization flags not work. 
Not fixed yet.

>   2) DBG = 0 and GCC 4 creates a load of new warnings
> that weren't there 
>  before. For example, I can't build ntoskrnl anymore. This is because 
>  it's detecting "uninitialized variables" which are actually being 
>  initalized by the macros themselves (see mm/pe.c and elf.inc.h). This 
>  did not happen before rbuild. 
FIXED.

>   3) Strip isn't being executed for dbg = 0
> anymore... I can add it but 
>  I have no idea how to make it dbg = 0 only. 
Not fixed. The ntoskrnl binary is 1.3MB instead of ~500KB. Due to #1 and 
also because strip -x isn't being executed.

>  
>  Not sure:
>  4) Is the OPTIMIZED envirovar still taken into account? I didn't seem 
>  to see so. 
Fixed (well, looks like we won't be supporting it, I don't mind)

>  
>  Best regards,
>  Alex Ionescu 
Best regards,
Alex Ionescu