Keith Owens a écrit :
> On Mon, 27 Nov 2000 16:02:13 -0600,
> > [Albert D. Cahalan]
> >> > Somebody else posted a reasonable hack for the [<>] problem. His
> >> > proposal involved letting multiple values share the same markers,
> >> > something like this:
> >Me too. (: Keith posed two objections:
> >
> >1. The >] could get word-wrapped so that it doesn't appear on the same
> > line as the [<. I *do not* see what makes this hard to parse
> > reliably.
>
> People seem to have forgotten that reading an oops from the screen is
> not the only source of data. Many oops are read from syslog which
> contains lots of different lines, most of which have no identification.
> ksymoops has to pick out oops text from a syslog and ignore all the
> non-oops lines.
>
When the oops is inside an interrupt, ther in no sync, and the only information
is on the 24 lines of the screen (not 25, because the oops ends with a "\n"
that kills one line at the top.
When you have a minor oops, you can add all information you want.
when you have a major oops that stops the machine,
implying a restart with fsck, you have only 25x80 characters.
So, 5 chars between 2 words is TOO MUCH !
why do not use only ONE special character as "~" "!" or ";"
instead of ">] [<"
>
> If the oops text is just a hex number with no identifying characters
> then it is very difficult to pick out oops text from all the other
> noise in syslog. ksymoops already gets false positives and prints some
> non-oops text, this confuses users who think that these lines are
> related to the oops.
>
> Removing [< >] increases the already high level of ambiguity and false
> positives in oops reporting from syslog. The presence of the marker
> characters makes the output more robust when line wrapped, without the
> markers a line wrapped trace is just a hex number.
yes, but use a marker made of only ONE character !
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Nov 30 2000 - 21:00:19 EST