Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Paul Wise
On Tue, 2019-05-14 at 13:57 +0100, Ben Hutchings wrote:

> Does user-mode-linux need to be a separate source package at all?

I guess it is separate to not require building yet another variant.

Looks like merging it was discussed in 2016:

https://bugs.debian.org/837920

-- 
bye,
pabs

https://wiki.debian.org/PaulWise



signature.asc
Description: This is a digitally signed message part


Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Paul Wise
Control: notforwarded -1
 
On Tue, 2019-05-14 at 14:36 +0530, Ritesh Raj Sarraf wrote:

> I submitted this fix upstream to the kernel some time ago. I don't
> recollect if the 4.19 kernel is what had it included.
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=9ca19a3a3e2482916c475b90f3d7fa2a03d8e5ed

That patch is already applied in the Debian user-mode-linux package:

debian/patches/fix-port-helper-path.patch

I noticed that the path embedded in the binary has an extra slash in it
compared to what the patch changes it to so I think the problem is that
the OS_LIB_PATH variable isn't defined correctly on amd64.

There seems to be some sort of confusion about the CONFIG_64BIT var but
it seems like the paths and variables in the sources are correct so I'm
not sure what is going on.

BTW, is it possible to do source-only uploads so that the build logs
are available on amd64, or do the versioned binaries mean no logs? If
so I think it would be good to only upload i386 binary packages so that
we get the build logs for the most used architecture.

~/linux-4.19.37 $ grep -rC1 OS_LIB_PATH 
arch/um/drivers/xterm.c-char *argv[] = { terminal_emulator, 
title_switch, title, exec_switch,
arch/um/drivers/xterm.c: OS_LIB_PATH 
"/uml/port-helper", "-uml-socket",
arch/um/drivers/xterm.c- file, NULL };
--
arch/um/include/shared/os.h-#ifdef CONFIG_64BIT
arch/um/include/shared/os.h:#define OS_LIB_PATH "/usr/lib64/"
arch/um/include/shared/os.h-#else
arch/um/include/shared/os.h:#define OS_LIB_PATH "/usr/lib/"
arch/um/include/shared/os.h-#endif
--
arch/um/os-Linux/main.c-
arch/um/os-Linux/main.c:#define UML_LIB_PATH":" OS_LIB_PATH "/uml"
arch/um/os-Linux/main.c-

~/user-mode-linux-4.19-1um $ grep -r CONFIG_64BIT
config.i386:# CONFIG_64BIT is not set
config.amd64:CONFIG_64BIT=y

-- 
bye,
pabs

https://wiki.debian.org/PaulWise



signature.asc
Description: This is a digitally signed message part


Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Ben Hutchings
On Tue, 2019-05-14 at 18:13 +0530, Ritesh Raj Sarraf wrote:
> On Tue, 2019-05-14 at 10:45 +0100, Ben Hutchings wrote:
> > > Irrespective, I think it is a fair candidate to be backported if
> > 4.19
> > > missed that window. For User-Mode-Linux, The sources are the same
> > that
> > > the Debian Linux kernel team maintains. So either they could pick
> > this
> > > commit as a patch  or else I'll try to look into it soon.
> > [...]
> > 
> > Can you please send this request to sta...@vger.kernel.org as well?
> 
> Thanks for the suggestion Ben.
> 
> The fix for this bug is now queued for the 4.19 stable tree. So soon,
> this fix will show up in the Debian Linux kernel package too. That is
> when we can trigger a rebuild of user-mode-linux package and then mark
> this bug fixed.
> 
> Debian Buster already has the dependency set correct to the 4.19
> kernel. Wishfully it would be much nicer if this whole process of the
> build could be automated to trigger periodically, or trigger on every
> upload of the linux-source package.

Does user-mode-linux need to be a separate source package at all?

It might be difficult to integrate it with the linux source package, as
we currently assume a strict separation between kernel and userland
binary packages, but it is probably possible to do.  Something to
consider for the bullseye release cycle.

Ben.

-- 
Ben Hutchings
I haven't lost my mind; it's backed up on tape somewhere.




signature.asc
Description: This is a digitally signed message part


Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Ritesh Raj Sarraf
On Tue, 2019-05-14 at 10:45 +0100, Ben Hutchings wrote:
> > Irrespective, I think it is a fair candidate to be backported if
> 4.19
> > missed that window. For User-Mode-Linux, The sources are the same
> that
> > the Debian Linux kernel team maintains. So either they could pick
> this
> > commit as a patch  or else I'll try to look into it soon.
> [...]
> 
> Can you please send this request to sta...@vger.kernel.org as well?

Thanks for the suggestion Ben.

The fix for this bug is now queued for the 4.19 stable tree. So soon,
this fix will show up in the Debian Linux kernel package too. That is
when we can trigger a rebuild of user-mode-linux package and then mark
this bug fixed.

Debian Buster already has the dependency set correct to the 4.19
kernel. Wishfully it would be much nicer if this whole process of the
build could be automated to trigger periodically, or trigger on every
upload of the linux-source package.


-- 
Ritesh Raj Sarraf | http://people.debian.org/~rrs
Debian - The Universal Operating System


signature.asc
Description: This is a digitally signed message part


Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Ben Hutchings
On Tue, 2019-05-14 at 14:36 +0530, Ritesh Raj Sarraf wrote:
> Control: tag -1 +fixed-upstream
> 
> CC: Adding Debian Kernel Maintainers
> 
> 
> Hello Paul,
> 
> I submitted this fix upstream to the kernel some time ago. I don't
> recollect if the 4.19 kernel is what had it included.

It went into 4.20 and hasn't been applied to 4.19-stable.

> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=9ca19a3a3e2482916c475b90f3d7fa2a03d8e5ed
> 
> Irrespective, I think it is a fair candidate to be backported if 4.19
> missed that window. For User-Mode-Linux, The sources are the same that
> the Debian Linux kernel team maintains. So either they could pick this
> commit as a patch  or else I'll try to look into it soon.
[...]

Can you please send this request to sta...@vger.kernel.org as well?

Ben.

-- 
Ben Hutchings
I haven't lost my mind; it's backed up on tape somewhere.




signature.asc
Description: This is a digitally signed message part


Re: Bug#928924: user-mode-linux: xterm functionality broken due to wrong path to port-helper

2019-05-14 Thread Ritesh Raj Sarraf
Control: tag -1 +fixed-upstream

CC: Adding Debian Kernel Maintainers


Hello Paul,

I submitted this fix upstream to the kernel some time ago. I don't
recollect if the 4.19 kernel is what had it included.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=9ca19a3a3e2482916c475b90f3d7fa2a03d8e5ed

Irrespective, I think it is a fair candidate to be backported if 4.19
missed that window. For User-Mode-Linux, The sources are the same that
the Debian Linux kernel team maintains. So either they could pick this
commit as a patch  or else I'll try to look into it soon.


Thanks,
Ritesh


On Mon, 2019-05-13 at 13:01 +0800, Paul Wise wrote:
> Package: user-mode-linux
> Version: 4.19-1um-1
> Severity: important
> File: /usr/bin/linux.uml
> 
> linux.uml in a GNOME terminal gives these errors in new tabs:
> 
> There was an error creating the child process for this terminal
> Failed to execute child process “port-helper” (No such file or
> directory)
> 
> The linux.uml binary references the wrong path for port-helper:
> 
> $ strings /usr/bin/linux.uml | grep port-helper
> /usr/lib//uml/port-helper
> 
> $ dpkg -L uml-utilities | grep port-helper
> /usr/lib64/uml/port-helper
> 
> Please fix this issue in Debian buster, it is an annoying bug.
-- 
Ritesh Raj Sarraf | http://people.debian.org/~rrs
Debian - The Universal Operating System


signature.asc
Description: This is a digitally signed message part