How does TERM=dumb support work?

[Carsten Mattner]

On Tue, Sep 25, 2012 at 6:17 PM, Augie Fackler <raf at durin42.com> wrote:
> On Tue, Sep 25, 2012 at 11:08 AM, Carsten Mattner
> <carstenmattner at gmail.com> wrote:
>> On Tue, Sep 25, 2012 at 3:07 PM, Augie Fackler <raf at durin42.com> wrote:
>>>
>>> On Sep 25, 2012, at 6:42 AM, Carsten Mattner <carstenmattner at gmail.com> wrote:
>>>
>>>
>>> What you're seeing actually strikes me as a bug: this makes anything
>>> with a progress bar unusable in emacs shell-mode, for example. Is
>>
>> Made another test and the hg progress bar seems not affected
>> by TERM=dumb at all and works the same way with TERM=9term.
>> The trick to make hg log behave is either to use a dumb PAGER
>> or omit some of the fancy /usr/bin/less options and most certainly
>> also disable the color extension.
>
> I'm not sure where log is coming from in this discussion. I thought we
> were talking about progress?

Sorry, it's the next test I did after cloning.

Either I'm confused or did test something else but while retesting
this again hg clone's progress bar didn't fit in 9term regardless of TERM=dumb.
Maybe some code was believing the terminal was wider than it is.

I must confess that I must have probably tested with the modified hgrc
before but now I had to use TERM=dumb so that progress did not emit
\r as I suppose that's the issue with 9term. IIRC 9term doesn not support \r.
That makes more sense in the context of the s390 patches.

>> Either way hg behaves better because you can configure ~/.hgrc
>> to not emit escape sequences in hg log while I failed to configure
>> the same for git.
>
> I'm not sure why you're seeing this. Is this pure plan9, or plan9 from
> user space on some other environment?

plan9port but I can also test in 9vx if I can get hg working there.
It's probably best if you do some independent testing.

>> The question is what's different in hg's progress extension that makes
>> hg clone not emit 9term-incomprehensible output.
>
> Progress just uses \r judiciously to overprint a line. It should still
> be disabled when TERM=dumb, and I've given myself a bug in bugzilla to
> track that.

Seems to be disabled. Without TERM=dumb it prints many many lines.

Another minor issue is that after "adding manifests" there's a blank
line and wrong indenting (# means space and #=# blank line):
[...]
adding manifest
#=#
#adding file changes
#=#
#added 3398 changesets[...]
#=#
#7494 files updates