umlbuilder-devel

[umlbuilder-devel] List of Bugs

[Jonathan R. <fre...@ul...>]

Hi Roger,

Here's a list of bugs and enhancement requests for you to fix that
were observed when running 1.4.0-5.

Bugs
----

0. (already reported this one) For umlbuilder 1.4.0-5 python 2.2
   requires that line 945 of gui.py has colum written as column.

        b.grid(row=2,colum=1, sticky=E)

		should be ..

        b.grid(row=2,column=1, sticky=E)

  Otherwise, the gui won't work.

1.  umlbuilder_gui is failing with this error:

    turbo indexer not available, falling back on traditional indexer
    Segmentation fault

However, the the cli version of umlbuilder doesn't fail with this
exception and proceeds to make the rootfs.  On my system it is falling
back to the traditonal indexer and causing this exception when the gui
is used.

2.  Although the rootfs was created, the initrd wasn't because the
    miniroot failed with this error:

Creating new initial ramdisk
Miniroot starting
mount: Mounting /proc on /proc failed: Bad address
mount: /proc/mounts: No such file or directory
mount: Mounting /dev on /dev failed: Bad address
grep: /proc/cmdline: No such file or directory
Miniroot exiting - exit status 1
System halted.

3.  When changing useinitrd="true" to useinitrd="false" in control,
    start still failed with this error:

Updating modules
Miniroot starting
mount: Mounting /proc on /proc failed: Bad address
mount: /proc/mounts: No such file or directory
mount: Mounting /dev on /dev failed: Bad address
grep: /proc/cmdline: No such file or directory
Miniroot exiting - exit status 1
System halted.

Obviously, the miniroot provided in this release needs some fixing.

Enhancement Requests
--------------------

1.  Add a --module-dir <dir> option and synonymous functionality to the gui
    which allows the user to choose an optional location to look for
    modules*tar.  Don't proceed until modules* are actually found,
    otherwise the whole process has to be done all over.

2.  Add a --uml-kernel <path> option and synonymous functionality to the
    gui which allows the user to choose an optional uml kernel


Thanks,


Jonathan

[umlbuilder-devel] Re: List of Bugs

[Roger B. <ro...@ro...>]

> 0. (already reported this one) For umlbuilder 1.4.0-5 python 2.2
>    requires that line 945 of gui.py has colum written as column.
> 
>         b.grid(row=2,colum=1, sticky=E)

Thanks for spotting the typo.  I am busy working on that one.

>     turbo indexer not available, falling back on traditional indexer
>     Segmentation fault

I'll look into this.  I normally do most of my testing with the
CLI version since I script it to run all the different distros.
The gui version effectively gathers up parameters and invokes
the cli code, except for the initial indexing.

> Creating new initial ramdisk
> Miniroot starting
> mount: Mounting /proc on /proc failed: Bad address
> mount: /proc/mounts: No such file or directory
> mount: Mounting /dev on /dev failed: Bad address
> grep: /proc/cmdline: No such file or directory
> Miniroot exiting - exit status 1
> System halted.

This is a bug in UML and skas.  It doesn't occur in TT mode.  Jeff
has fixed it in the most recent UML release (a day or two ago).

> 3.  When changing useinitrd="true" to useinitrd="false" in control,
>     start still failed with this error:
> 
> Updating modules
> Miniroot starting
> mount: Mounting /proc on /proc failed: Bad address
> mount: /proc/mounts: No such file or directory
> mount: Mounting /dev on /dev failed: Bad address
> grep: /proc/cmdline: No such file or directory
> Miniroot exiting - exit status 1
> System halted.

Same problem.  miniroot is trying to install modules instead of mkinitrd

> Obviously, the miniroot provided in this release needs some fixing.

For once it isn't my fault :-)

There are actually some nice goodies in miniroot.  As well as making
initial ramdisks and installing modules, it has a rescue mode that
gets you a command line shell.  rescue mode doesn't premount your
filesystems yet.

It also has a filesystem resizer/format changer.  This works by
making a new sparse file, running mkfs, and then cpio the data
across.

> 1.  Add a --module-dir <dir> option and synonymous functionality to the gui
>     which allows the user to choose an optional location to look for
>     modules*tar.  Don't proceed until modules* are actually found,
>     otherwise the whole process has to be done all over.

Will do.
 
> 2.  Add a --uml-kernel <path> option and synonymous functionality to the
>     gui which allows the user to choose an optional uml kernel

At runtime, this is easy.  The control script looks in its directory for
an executable with the same name as the umid.  It then looks for one
named linux.  Consequently you can just copy or symlink in the one you
want.

That can only happen after install so I'll also add preinstall hooks.

Thanks,

Roger

[umlbuilder-devel] Re: List of Bugs

[Jonathan R. <fre...@ul...>]

At Tue, 4 Feb 2003 16:19:00 -0800,
Roger Binns wrote:

> This is a bug in UML and skas.  It doesn't occur in TT mode.  Jeff
> has fixed it in the most recent UML release (a day or two ago).

I hadn't noticed this.  Tested it out with the newest patch, and works
fine with only one minor error observed that produces no side effects:

Miniroot starting
Miniroot exiting - exit status 0
Couldn't umount /dev on none: Bad address
Couldn't umount /dev on none: Bad address
System halted.

Creating new initial ramdisk
Checking for the skas3 patch in the host...found

Checking for /proc/mm...found
Miniroot starting
174+0 records in
174+0 records out
mke2fs 1.27 (8-Mar-2002)
Miniroot exiting - exit status 0
Couldn't umount /dev on none: Bad address
Couldn't umount /dev on none: Bad address
System halted

There's just a problem with umounting /dev.  The question is, should
/dev be umounted anyways?

Another enhancement request:  In umlbuilder help add more explanation
for 'list'.

(*) You can supply 'list' to this argument to get a list of possible values
    ex: umlbuilder --distro list
	umlbuilder --distro d --modules list

The reason why I suggest this is because it's intuitive to want to do
something like

umlbuilder --modules list

which fails unless --distro d is included, and ofcourse user's don't
always looks closely at the documentation.

> There are actually some nice goodies in miniroot.  As well as making
> initial ramdisks and installing modules, it has a rescue mode that
> gets you a command line shell.  rescue mode doesn't premount your
> filesystems yet.
> 
> It also has a filesystem resizer/format changer.  This works by
> making a new sparse file, running mkfs, and then cpio the data
> across.

Nice stuff.  Eventually - future project - I will add umlbuilder to
gbootroot using the CLI interface, though the GUI could be used, too.
I'm still deciding whether or not to implement a plug-in system, or
just to continue building things directly into gbootroot.  Umlbuilder
definitely fits into gbootroot's IDE design. 

Thanks,


Jonathan