RE: Bug#55820: apt grossly misestimates download
> -----Original Message-----
> From: Jason Gunthorpe [mailto:jgg@ualberta.ca]
> Sent: Friday, January 21, 2000 1:37 PM
> To: Ross Boylan
> Cc: 55820@bugs.debian.org; APT Development Team
> Subject: RE: Bug#55820: apt grossly misestimates download
>
>
>
> On Fri, 21 Jan 2000, Ross Boylan wrote:
>
> > OK, what more information do you need?
> >
> > Considering that I sat in front of the computer as it listed
> files it was
> > downloading, and that it went through them slowly (I have a
> modem), I think
>
> Are you saying it continued to print things on the screen even after the
> progress meter 'disappeared'
Yes
> and what exactly do you mean by disappeared.
The progress meter was, as I recall, on the lower right corner of the
display. It went up to 99% and then vanished--i.e., just blank screen
there. The program definitely continued to print things (and download)
after this point.
> Did you ignore any telling error messages?
>
Yes (depending on your definition of telling). Every time I run apt I get
lots of error messages, some of which may relate to the situation I
described in my initial report and later in this note (the messages things
like unable to find repository, as well as other more serious-sounding ones.
I can't recall the specifics, and would love to know the best way to capture
them). I ignored them in the sense that I kept going.
> There is one way I can see what you describe happening, but for that to
> happen the progress meter could not go away.
>
> > your statement that "there certainly is no more downloading
> going on" is off
> > base. Certainly closing a bug because you think it can't
> happen isn't good
> > practice.
>
> I have a general rule, the more obvios the bug (and the more unlikely it
> is to happen) the better chance it is user-error and not a programming
> problem.. This is *REALLY* obvios and *REALLY* unlikely :>
That's not a bad rule, but given the state of flux of things right now,
perhaps not as reliable as usual. Also, if the system is such that a user
following it through is led to make an error, it really is a sign the system
needs to be changed even if the problem is not a bug.
>
> For instance, you may have your screen/console misconfigured which gives
> the effect of a vanishing progress bar.
Well, there are problems with libncurses and slang in the boot-floppy 2.2.4
incarnation of the system. But since the progress meter was showing
numbers, and they did reach 99%, I have a feeling this particular problem
may lie elsewhere.
Also, I'm pretty sure that it said it was going to get 6 packages, and then
got 9. It may have been after 6 that the overall progress bar stopped.
(Numbers approximate).
>
> Jason
>
By the way, there were a lot of other ways in which dselect/apt-get seemed
to go wrong, but I haven't had a chance to check on the details. In
particular, I think it missed some files I have on cd.
It would be really helpful to know the best way to capture the messages that
fly by during a session, so I and you have something more definite to go on.
One problem is that apt problems are tricky to reproduce, since one's system
often ends up in a different configuration at the end of a session.
Also, if the thing I am trying to do is just something that shouldn't work,
let me know. I'll stop trying to do it! I downloaded a bunch of .deb
files, some to my local hard drive and some burned to CD. I preserved the
directory structure of the original archive. However, control files like
Packages are either missing or from a different date than the debs. Even if
they are present, they obviously don't reflect the fact that I only have
some of the debs (and sources) in the archive. Someone on debian-user told
me that apt would be able to use the files in this situation. This is also
why I have expected a certain amount of complaining from it about missing
files.
P.S. I've left the apt development team on the cc list, but I'm a little
concerned they'll get duplicates via the 55820@bugs. I'll be happy to trim
the list if that's a problem.
Reply to: