smart upgrade crashes on serial or xen terminal

[Rehan Khan]

Imported: http://tracker.labix.org/issue172
http://tracker.labix.org/issue355

Reason for Import: Issue Review Patch review

further details: https://blueprints.launchpad.net/smart/+spec/bug-reporting-migration

[Rehan Khan (rasker)]

Issue 172:

msg1276 (view) Author: netmask Date: 2008-01-04.11:50:00

eswierk, thanks for taking a look at this issue. Even though looks like you
touched the correct spot to fix the issue, IMHO one should not just guess that
user is running on a 70px window. My vote would be 20px as a safe value.

msg1269 (view) Author: eswierk Date: 2007-12-31.18:55:26

When TextProgress.setScreenWidth is passed a width of 0, it generates a bogus
format string containing a negative number.

The attached patch works around this bug by forcing the width to at least 70
characters.

msg1185 (view) Author: ed Date: 2007-08-25.14:54:44

NB I can still reproduce this on Fedora 7.

msg1131 (view) Author: ed Date: 2007-04-08.19:32:31

Hi, yes I can still reproduce this on Fedora Core 5 with the default set of
repositories. However, it seems to appear only when running smart inside an
Emacs shell buffer. (This is a way to run a subshell in Emacs and have the
output of commands appear in Emacs.) Try starting GNU Emacs or xemacs and do
M-x shell to make the shell buffer appear, then run a smart command to install a
package. My hunch is that the progress bar code is going wrong because smart's
output is not going to a real tty that supports querying width and height. Here
is another example:

root@bone /home/ed/packages/SRPMS # smart install dosbox
Loading cache...
( 0%) [100%]

Computing transaction...

Installing packages (2):

736.1kB of package files are needed. 2.1MB will be used.

Confirm changes? (Y/n): y

Fetching packages...
-> http://download.fedora.redhat.com/.../SDL_net-1.2.6-2.fc6.x86_64.rpm
( 0%)(73% Stalled Unknown)(73% 4.0kB/s Unknown) [ 50%]
-> http://download.fedora.redhat.com/.../dosbox-0.70-1.fc6.x86_64.rpm
(4% 4.0kB/s 02m55s)(4% 7.0kB/s 02m55s)(6% 7.0kB/s 01m37s)(8% 7.0kB/s 01m37s)(8%
13.2kB/s 01m37s)(11% 13.2kB/s 49s)(13% 13.2kB/s 49s)(13% 17.9kB/s 49s)(15%
17.9kB/s 34s)(17% 21.4kB/s 34s)(20% 21.4kB/s 27s)(22% 24.1kB/s 27s)(24% 24.1kB/s
23s)(26% 24.1kB/s 23s)(26% 26.0kB/s 23s)(29% 26.0kB/s 20s)(29% 23.5kB/s 20s)(31%
23.5kB/s 21s)(33% 23.5kB/s 21s)(33% 25.6kB/s 21s)(35% 25.6kB/s 18s)(38% 25.6kB/s
18s)(38% 27.2kB/s 18s)(40% 27.2kB/s 16s)(42% 27.2kB/s 16s)(42% 28.4kB/s 16s)(44%
28.4kB/s 14s)(47% 28.4kB/s 14s)(47% 29.3kB/s 14s)(49% 29.3kB/s 12s)(51% 29.3kB/s
12s)(51% 30.0kB/s 12s)(53% 30.0kB/s 11s)(55% 30.5kB/s 11s)(58% 30.5kB/s 10s)(60%
30.5kB/s 10s)(60% 30.9kB/s 10s)(62% 30.9kB/s 09s)(64% 31.1kB/s 09s)(67% 31.1kB/s
08s)(67% 27.3kB/s 08s)(69% 27.3kB/s 08s)(71% 27.3kB/s 08s)(71% 28.5kB/s 08s)(73%
28.5kB/s 07s)(76% 28.5kB/s 07s)(76% 29.4kB/s 07s)(78% 29.4kB/s 05s)(80% 29.4kB/s
05s)(80% 30.0kB/s 05s)(82% 30.0kB/s 04s)(85% 30.0kB/s 04s)(85% 30.5kB/s 04s)(87%
30.5kB/s 03s)(89% 30.5kB/s 03s)(89% 30.9kB/s 03s)(91% 30.9kB/s 02s)(94% 31.2kB/s
02s) (96% 31.2kB/s 01s) (98% 31.2kB/s 01s) (98% 31.4kB/s 01s) [100%]

Committing transaction...
( 0%) [ 0%]

Traceback (most recent call last):
  File "/usr/bin/smart", line 194, in ?
    main(sys.argv[1:])
  File "/usr/bin/smart", line 167, in main
    exitcode = iface.run(opts.command, opts.argv)
  File "/usr/lib64/python2.4/site-packages/smart/interface.py", line 53, in run
    result = _c...

[Rehan Khan (rasker)]

msg1293 (view) Author: peter-endian Date: 2008-01-28.11:01:11

yes, that patch fixes the problem also. Did not see it before.
However, viewing the patch i think, the patch of issue172 is more a workaround
only for this specific problem.

But the problem is more that getScreenWidth sometimes can return 0 and that size
will be used also in other parts of the tool, which will also fail if fixed only
in text/progress.py like in patch of issue172

I think it would be better to fix it in getScreenWidth(). I am unsure if simply
returning 80 if the size is 0 is sufficient. Better would be to read it out
somewhere, but did not find any possibility for serial terminal emulation.
Also maybe returning 80 if the width is <20, like you mentioned would be better,
than simply returning 80 if it is 0, since the problem persists if the width
would be 4 for example.

msg1291 (view) Author: netmask Date: 2008-01-27.11:22:23

Hi Peter,
Could you please check if the proposed patch in issue172 fixes your issue?

msg1289 (view) Author: peter-endian Date: 2008-01-25.15:24:40

if you try to smart upgrade through a serial console or a xen console smart
crashes with this traceback:

...
  File "/usr/lib/python2.4/site-packages/smart/backends/rpm/pm.py", line 309, in
__call__
    self.prog.show()
  File "/usr/lib/python2.4/site-packages/smart/progress.py", line 110, in show
    self.expose(topic, percent, *info)
  File "/usr/lib/python2.4/site-packages/smart/interfaces/text/progress.py",
line 108, in expose
    out.write(self._topicmaskn % (n, topic))
ValueError: unsupported format character '-' (0x2d) at index 9

This happens because getScreenWith() returns a window size of 0 for those
terminal emulations.
The attached patch returns a default value of 80 when the window size is 0.

[Rehan Khan (rasker)]

The above is issue 355