rbuild status - need help finalizing

[Royce Mitchell III]

rbuild has one bug ( that we're aware of ) blocking it from being ready 
to go live.

Basically, both bootcd and livecd are halting with a 
"inaccessible_boot_device" after blue.sys.

Would anybody be willing to look at it and help us figure out what we're 
doing wrong when generating the ISOs?

Also, being this close, would everybody who cares please check out the 
rbuild branch for feedback before we go live?

Here's the branch url: svn://svn.reactos.com/branches/xmlbuildsystem/reactos

Thanks

[WaxDragon]

I still can't build this branch:

[RSYM] output-i386/subsys/system/ibrowser/ibrowser.exe
[DLLTOOL] output-i386/lib/fmifs/fmifs.a
[CC] subsys/system/format/format.c
subsys/system/format/format.c: In function `wmain':
subsys/system/format/format.c:476: warning: int format, different type
arg (arg 2)
subsys/system/format/format.c:477: warning: int format, different type
arg (arg 2)
make: *** [obj-i386/subsys/system/format/format.o] Error 1

Which is "%I64d" on the printf, this works in trunk.

I build under linux with "gcc version 3.4.2 (mingw-special)"

Here are the env vars I've been setting manually in HEAD:
 export PATH=$PATH:$HOME/bin/mingw/bin
 export HOST=mingw32-linux
 export ROS_INSTALL=$HOME/vmware/ReactOS/ros/ReactOS
 export INSTALL_DIR=$ROS_INSTALL
 export ROS_USE_PCH=yes

On 4/23/05, Royce Mitchell III [e-mail] wrote:
> rbuild has one bug ( that we're aware of ) blocking it from being ready
> to go live.
>
> Basically, both bootcd and livecd are halting with a
> "inaccessible_boot_device" after blue.sys.
>
> Would anybody be willing to look at it and help us figure out what we're
> doing wrong when generating the ISOs?
>
> Also, being this close, would everybody who cares please check out the
> rbuild branch for feedback before we go live?
>
> Here's the branch url: svn://svn.reactos.com/branches/xmlbuildsystem/reactos
>
> Thanks
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev
-- 
"<tinus_> and also win32k should be fixed up a bit"

[Casper Hornstrup]

Please svn up and try again
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of WaxDragon
> Sent: 23. april 2005 18:55
> To: ReactOS Development List
> Subject: Re: [ros-dev] rbuild status - need help finalizing
>
> I still can't build this branch:
>
> [RSYM] output-i386/subsys/system/ibrowser/ibrowser.exe
> [DLLTOOL] output-i386/lib/fmifs/fmifs.a
> [CC] subsys/system/format/format.c
> subsys/system/format/format.c: In function `wmain':
> subsys/system/format/format.c:476: warning: int format, different type
> arg (arg 2)
> subsys/system/format/format.c:477: warning: int format, different type
> arg (arg 2)
> make: *** [obj-i386/subsys/system/format/format.o] Error 1
>
> Which is "%I64d" on the printf, this works in trunk.
>
> I build under linux with "gcc version 3.4.2 (mingw-special)"
>
> Here are the env vars I've been setting manually in HEAD:
>  export PATH=$PATH:$HOME/bin/mingw/bin
>  export HOST=mingw32-linux
>  export ROS_INSTALL=$HOME/vmware/ReactOS/ros/ReactOS
>  export INSTALL_DIR=$ROS_INSTALL
>  export ROS_USE_PCH=yes

[WaxDragon]

Now I get this...

[CC] subsys/system/regedit/treeview.c
[WRC] obj-i386/subsys/system/regedit/rsrc.coff
[LD] output-i386/subsys/system/regedit/regedit.exe
[RSYM] output-i386/subsys/system/regedit/regedit.exe
[CC] subsys/system/regsvr32/regsvr32.c
[WRC] obj-i386/subsys/system/regsvr32/regsvr32.coff
[LD] output-i386/subsys/system/regsvr32/regsvr32.exe
[RSYM] output-i386/subsys/system/regsvr32/regsvr32.exe
[CC] subsys/system/reporterror/reporterror.c
[WRC] obj-i386/subsys/system/reporterror/reporterror.coff
subsys/system/reporterror/reporterror.rc:19:17: Dk.rc: No such file or directory
make: *** [obj-i386/subsys/system/reporterror/reporterror.coff] Error 1

WD
-- 
"<tinus_> and also win32k should be fixed up a bit"

[Casper Hornstrup]

Filename casing error. Fixed in svn now.

Casper
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of WaxDragon
> Sent: 23. april 2005 19:45
> To: ReactOS Development List
> Subject: Re: [ros-dev] rbuild status - need help finalizing
>
> Now I get this...
>
> [CC] subsys/system/regedit/treeview.c
> [WRC] obj-i386/subsys/system/regedit/rsrc.coff
> [LD] output-i386/subsys/system/regedit/regedit.exe
> [RSYM] output-i386/subsys/system/regedit/regedit.exe
> [CC] subsys/system/regsvr32/regsvr32.c
> [WRC] obj-i386/subsys/system/regsvr32/regsvr32.coff
> [LD] output-i386/subsys/system/regsvr32/regsvr32.exe
> [RSYM] output-i386/subsys/system/regsvr32/regsvr32.exe
> [CC] subsys/system/reporterror/reporterror.c
> [WRC] obj-i386/subsys/system/reporterror/reporterror.coff
> subsys/system/reporterror/reporterror.rc:19:17: Dk.rc: No such file or directory
> make: *** [obj-i386/subsys/system/reporterror/reporterror.coff] Error 1
> WD
>
> "<tinus_> and also win32k should be fixed up a bit"
>
> Ros-dev mailing list
> [e-mail]
> http://reactos.com:8080/mailman/listinfo/ros-dev

[WaxDragon]

Thanks to Casper for fixing the couple of bug I was seeing during
building.  Now down to the meat of it.  I compared the bootcd from
rbuild and the bootcd from trunk using 'isoinfo' from my slackware
box.  Here are two listings of the guts of each, and dumps of the path
tables from both.  There are several differences...

Included are files I generated, useful for doing a diff. ;0)

WD
-- 
"<tinus_> and also win32k should be fixed up a bit"

[Ge van Geldorp]

> From: Royce Mitchell III
> Sent: Saturday, April 23, 2005 16:43
> To: ReactOS Development List
> Subject: [ros-dev] rbuild status - need help finalizing
>
> rbuild has one bug ( that we're aware of ) blocking it from 
> being ready to go live.
>
> Basically, both bootcd and livecd are halting with a 
> "inaccessible_boot_device" after blue.sys.

It seems that a lot of drivers, including atapi.sys and cdrom.sys, don't
contain relocation information. You can check that using "objdump -p
cdrom.sys", which will show:

Characteristics 0x2103
        relocations stripped
        executable
        32 bit words
        DLL

Doing an "objdump -h cdrom.sys" also reveals there's no .reloc section.
This will cause the test at line 1200 of ntoskrnl/ldr/loader.c (function
LdrPEPerformRelocations()) to fail, which in turn means the driver is not
initialized. In the end, that seems to be responsible for the
INACCESSIBLE_BOOT_DEVICE.
I've tried removing the test from loader.c but that causes a page fault when
the driver entry point is called (which didn't really surprise me). The
problem thus seems to be that the relocation information is AWOL.

Gé van Geldorp.

[WaxDragon]

After getting a good look at the data I generated, I see that the ISO
itself seems fine, as Ge has pointed out.  So I went ahead and tried
to build KDB and DBG to get some more info, and I get this error:

[CC] drivers/video/displays/vga/objects/copybits.c
drivers/video/displays/vga/objects/copybits.c:4:1: "DBG" redefined
<command line>:3:1: this is the location of the previous definition
make: *** [obj-i386/drivers/video/displays/vga/objects/copybits.o] Error 1

WD
-- 
"<tinus_> and also win32k should be fixed up a bit"

[Ge van Geldorp]

> From: Royce Mitchell III
>
> Also, being this close, would everybody who cares please 
> check out the rbuild branch for feedback before we go live?

First, congratulations, I'm impressed by the work you and Casper have done.

I had one major problem (I need the *.nostrip.* files for GDB) which Casper
solved within two hours after I mentioned it to him. Again very impressive,
thanks.

There's two small issues, nice-to-have but not essential as far as I'm
concerned. First is that used to be able to do "make" followed by "make
install" from a subdirectory. The "make install" doesn't work anymore ("No
rule to make target `install'."). The other thing is that I used to be able
to setup a bootcd using "make bootcd", then I would go in, recompile
something and copy it manually to ..\bootcd\disk\wherever. After that I
could do a "make bootcd_makecd" so it would just use the files in
..\bootcd\disk to create the .iso, without having to go through the whole
reactos tree. "make bootcd" is now one big target so that doesn't work
anymore.
I can probably fix both issues myself, but since you asked for feedback and
I'm lazy...

Gé van Geldorp.

[Casper Hornstrup]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of Ge van Geldorp
> Sent: 23. april 2005 23:23
> To: 'ReactOS Development List'
> Subject: [ros-dev] rbuild feedback
> > From: Royce Mitchell III
> >
> > Also, being this close, would everybody who cares please
> > check out the rbuild branch for feedback before we go live?
>
> First, congratulations, I'm impressed by the work you and Casper have done.
>
> I had one major problem (I need the *.nostrip.* files for GDB) which Casper
> solved within two hours after I mentioned it to him. Again very impressive,
> thanks.

Thanks.
> There's two small issues, nice-to-have but not essential as far as I'm
> concerned. First is that used to be able to do "make" followed by "make
> install" from a subdirectory. The "make install" doesn't work anymore ("No
> rule to make target `install'.").

I'll fix that.
> The other thing is that I used to be able
> to setup a bootcd using "make bootcd", then I would go in, recompile
> something and copy it manually to ..\bootcd\disk\wherever. After that I
> could do a "make bootcd_makecd" so it would just use the files in
> ..\bootcd\disk to create the .iso, without having to go through the whole
> reactos tree. "make bootcd" is now one big target so that doesn't work
> anymore.

The issue here is wether or not to clean the bootcd directory before copying
the files into it. If we don't need to clean the directory first then we can
have 'make bootcd' copy the files using dependencies which will be quite fast
the second time since only the changed files will be copied. If we need to
clean the directory first then all files must always be copied into the bootcd
directory. The problem with not cleaning the directory first is that the
bootcd directory may be inconsistent due to bootcds made previously from other
revisions. 'make livecd' have the same issue. I could go either way so I'll
let you decide.
> I can probably fix both issues myself, but since you asked for feedback and
> I'm lazy...
>
> Gé van Geldorp.

[Jiri Doubek]

Hi!

Casper Hornstrup napsal(a):
> The issue here is wether or not to clean the bootcd directory before copying
> the files into it. If we don't need to clean the directory first then we can
> have 'make bootcd' copy the files using dependencies which will be quite fast
> the second time since only the changed files will be copied. If we need to
> clean the directory first then all files must always be copied into the bootcd
> directory. The problem with not cleaning the directory first is that the
> bootcd directory may be inconsistent due to bootcds made previously from other
> revisions. 'make livecd' have the same issue. I could go either way so I'll
> let you decide.

So, why not make in destination directory one subdir called with actual 
SVN revision number? And "make clean_bootcd_dirs" can clean those 
directories. This can solve the problem...

Jirka D.

[Casper Hornstrup]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of Jiri Doubek
> Sent: 24. april 2005 13:44
> To: ReactOS Development List
> Subject: Re: [ros-dev] rbuild feedback
>
> Hi!
>
> Casper Hornstrup napsal(a):
> > The issue here is wether or not to clean the bootcd directory before copying
> > the files into it. If we don't need to clean the directory first then we can
> > have 'make bootcd' copy the files using dependencies which will be quite fast
> > the second time since only the changed files will be copied. If we need to
> > clean the directory first then all files must always be copied into the bootcd
> > directory. The problem with not cleaning the directory first is that the
> > bootcd directory may be inconsistent due to bootcds made previously from other
> > revisions. 'make livecd' have the same issue. I could go either way so I'll
> > let you decide.
>
> So, why not make in destination directory one subdir called with actual
> SVN revision number? And "make clean_bootcd_dirs" can clean those
> directories. This can solve the problem...
>
> Jirka D.

Then you would need to manually delete the bootcd directories for the revisions.

Casper