bug-bash
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Possible bug in 'unset' builtin


From: Chet Ramey
Subject: Re: Possible bug in 'unset' builtin
Date: Mon, 11 May 2015 11:12:11 -0400
User-agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.10; rv:31.0) Gecko/20100101 Thunderbird/31.6.0

On 5/8/15 12:42 PM, Eduardo A. Bustamante López wrote:
>>> [19:24] < dualbus> for every fix that Chet does, he introduces like 2 bugs.
>>> Nice way of keeping himself busy
>>
>> That points to the importance of having a good, comprehensive test suite.
>> It's difficult to test the interaction between features otherwise.
> 
> I apologize for my comment :-) I sometimes do not acknowledge the amount of
> effort that you put into bash. It's amazing, and I'm very grateful for that.

No problem.  You do you.

> I do believe that the tests are not good enough though. The reasons:
> 
> 1. They're slow (I do not know how this could be fixed)

The job control test is the slowest one, and it's because it calls wait
on some sleep processes.  You can look at speeding that one up.

> 2. They're not automated (you have to eyeball them to see if there are false
> positives)

How would you like to automate them other than comparing expected output
to actual output in some way?  If you have ideas, please bring them to
the table.

> 3. There is not enough coverage (though this is easy to fix, with enough
> effort)

As I wrote in a response to another message, the coverage holes, such as
they are, primarily exist in the interactive aspects of the shell.  It's
difficult to test readline in an automated fashion.

> 4. They're not taken with enough seriousness (I ran these in freebsd and
> openbsd, and as of this date, multiple unicode tests are failing). IMO the
> tests should be run at least for linux/bsd before pushing the code to the
> public repository.

I run the tests on my primary development environment -- Mac OS X -- and
sometimes on Linux before pushing the devel branch updates.  The devel
branch is just that -- development.  It's not a final release, and I
don't expect it to be polished like one.

The unicode tests are tricky, too.  If you can see a way to make the BSDs
pass the failing (French) unicode/iso8859-1 tests without attributing it
to the BSDs just handling that locale in a different way, or missing the
locale in general, please let me know.

Chet

-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, ITS, CWRU    chet@case.edu    http://cnswww.cns.cwru.edu/~chet/



reply via email to

[Prev in Thread] Current Thread [Next in Thread]