first effort with git

first effort with git

[Ray Andrews]

I just tried a 'git'. It looks like it all came down the wire,
but on trying to build I get this:

.....

./B06fc.ztst: starting.
Pattern match failed:
<*%*
 >HP-y5--5-Debian1# HP-y5--5-Debian1# HP-y5--5-Debian1#
Test ./B06fc.ztst failed: error output differs from expected as shown 
above for:
   $ZTST_testdir/../Src/zsh +Z -fsi <<< $'fc -p /dev/null 0 0\n:'
Was testing: Checking that fc -p doesn't core dump when history size is zero
./B06fc.ztst: test failed.

.....

A rebuild of my virgin 5.0.7 shows no errors.

Re: first effort with git

[Peter Stephenson]

On Tue, 11 Nov 2014 19:37:53 -0800
Ray Andrews <rayandrews@eastlink.ca> wrote:
> I just tried a 'git'. It looks like it all came down the wire,
> but on trying to build I get this:
> 
> .....
> 
> ./B06fc.ztst: starting.
> Pattern match failed:
> <*%*
>  >HP-y5--5-Debian1# HP-y5--5-Debian1# HP-y5--5-Debian1#
> Test ./B06fc.ztst failed: error output differs from expected as shown 
> above for:
>    $ZTST_testdir/../Src/zsh +Z -fsi <<< $'fc -p /dev/null 0 0\n:'
> Was testing: Checking that fc -p doesn't core dump when history size is zero
> ./B06fc.ztst: test failed.

The test is equivalent to running

zsh +Z -fsi <<< $'fc -p /dev/null 0 0\n:'

and expecting to see a command line prompt, which would have a "%" in
it --- the "<*%*" hieroglyphic says "look at standard output from
the test and do a pattern match against the pattern '%*'".

Instead, somehow you're managing to get the shell to produce the output
"HP-y5--5-Debian1# HP-y5--5-Debian1# HP-y5--5-Debian1#" before
the command line shows up.  Quite likely something run from /etc/zshenv,
at least there's nothing much else it could be.

I suppose the test could be a bit more sophisticated about this: it just
needs to prove you got the command line prompt eventually.

pws

Re: first effort with git

[Peter Stephenson]

On Wed, 12 Nov 2014 10:38:22 +0000
Peter Stephenson <p.stephenson@samsung.com> wrote:
> Instead, somehow you're managing to get the shell to produce the output
> "HP-y5--5-Debian1# HP-y5--5-Debian1# HP-y5--5-Debian1#" before
> the command line shows up.  Quite likely something run from /etc/zshenv,
> at least there's nothing much else it could be.

Ah, I'm being stupid:  clearly

HP-y5--5-Debian1# 

*is* the prompt.  /etc/zshenv or something called from there is changing
it from the default and so it's failing to match "%*".

pws

Re: first effort with git

[Peter Stephenson]

On Wed, 12 Nov 2014 09:44:39 -0800
Ray Andrews <rayandrews@eastlink.ca> wrote:
> Default prompt is:
> 
>      HP-y5--5-Debian1#
> 
> ... but is that the default prompt? I just commented out the 'PS1' line 
> above.

I'm guessing that's set in /etc/zshenv.  That gets run for all zsh's
that aren't emulating something else.  So it looks like a default.

While that's not typical, it's probably something the test should cope
with, so if I've understood properly I don't think it's your problem...

The test is simply checking the shell doesn't dump core.  So probably
the right fix is to print a message after the "fc", then read input
until that's seen.  Hmmm... actually, that doesn't need even need to go
to stdout.  Suppose we redirect stdout from that test to /dev/null,
print the new message to stderr, and test that?

> I see that 'Test/B06fc.ztest' was changed. I pulled in the copy from 
> 5.0.7, and the
> error disappears. That's with my usual environment.

You mean, it's not there in an older version?  That's not a surprise,
the test is new.

pws

Re: first effort with git

[Bart Schaefer]

On Nov 12,  9:44am, Ray Andrews wrote:
}
}      HP-y5--5-Debian1#
} 
} ... but is that the default prompt? I just commented out the 'PS1' line 
} above.

The default prompt is '%m%# ' which means it ends with a % for a normal
user or with a # for root.

} I see that 'Test/B06fc.ztest' was changed. I pulled in the copy from 
} 5.0.7, and the error disappears.

That's because rolling back that file entirely removes the new tests that
examine the interactive prompt.

Re: first effort with git

[Bart Schaefer]

On Nov 12,  4:52pm, Peter Stephenson wrote:
}
} The test is simply checking the shell doesn't dump core.  So probably
} the right fix is to print a message after the "fc", then read input
} until that's seen.  Hmmm... actually, that doesn't need even need to go
} to stdout.  Suppose we redirect stdout from that test to /dev/null,
} print the new message to stderr, and test that?

That'll work for that particular test, but not for the two that come
after it which actually need to see the error text returned by "fc"
and which will have the same problem.

Re: first effort with git

[Ray Andrews]

On 11/12/2014 02:45 AM, Peter Stephenson wrote:


Peter, see below:
>
> Instead, somehow you're managing to get the shell to produce the output
> "HP-y5--5-Debian1# HP-y5--5-Debian1# HP-y5--5-Debian1#" before
> the command line shows up.  Quite likely something run from /etc/zshenv,
> at least there's nothing much else it could be.
> Ah, I'm being stupid:  clearly
>
> HP-y5--5-Debian1#
>
> *is* the prompt.  /etc/zshenv or something called from there is changing
> it from the default and so it's failing to match "%*".
>
> pws
>
Here's my prompt line:

    # THE PROMPT:
    #         yellow:tty       red:host         green:user blue:path
    PS1=$'\n%{\e[0;33;1m%}%l %{\e[0;31;1m%}%m %{\e[0;32;1m%}%n
    %{\e[0;34;1m%}%d $ %{\e[0m%}'


Producing this:

     pts/7 HP-y5--5-Debian1 root /aMisc/zsh-5.0.7/Src $


Default prompt is:

     HP-y5--5-Debian1#

... but is that the default prompt? I just commented out the 'PS1' line 
above.

I see that 'Test/B06fc.ztest' was changed. I pulled in the copy from 
5.0.7, and the
error disappears. That's with my usual environment.