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

Re: how to avoid stupid bugs?



On Sun, Feb 02, 2003 at 08:28:56PM +0200, Marius Andreiana wrote:
> In every first new red hat beta there are lots of stupid little bugs
> which shouldn't be there. I don't blame anybody, I'd like to discuss
> about this instead, maybe we can improve something.
> 
> Some recent examples : 
> * making a boot disk during install doesn't work

This is most likely due to a limitation in the Linux kernel's boot
sector (which normally is appended to the beginning of the vmlinuz
image).  There is a 1 MB limitation in the old boot sector on the size
of the kernel.  Peter Anvin posted new bootsector code for 2.5 that
overcomes this limit for floppies.  This is a hard problem, and it
wasn't something we could fix in time for the beta.  It was known to
be broken, but didn't make it in the release notes in time, since we
have to send them to translation.

> * kudzu and mice are completely messed up in phoebe2. Not detecting mice
> locked keyboard, and most of ps/2 and serial mice weren't detected. 

The PS/2 probe code was definitely broken in Phoebe beta 1.  It should
be much better in beta 2.  Do you have more hardware details?  The
serial probe has changed a bit, it is possible we don't get the probe
right on your serial mouse.

> * right clicking on a desktop launcher and selecting Properties brings
> up icon properties, not launcher properties

Might be a bug.  I'll let Havoc or Alex comment on that one.

> * xmms crashes on trying to play a non-existent file in play list

Yea, xmms is a little crashy sometimes.

> The problem with these bugs is that they appear out of nowhere, on
> features that used to work fine and that are very common use cases.
> There should be bugs only in new features. So why it this happening? 

Software is always changing, especially if upstream maintainers are
actively modifying features etc.  This is why we have a beta.  File
bugs on what you've found and we'll work on fixing them

> A reason I can think of is that major features are often (maybe too
> often) rewritten from scratch. It's not unusual to see on freshmeat
> announcements: "What's new: completely rewritten". How can a developer
> throw away tested and working code to start all over again, introducing
> new bugs which annoy users and waste testers time and theirs ( developer
> must fix these bugs, qa has to verify the fix it's working...). Maybe
> it's easier to write your own code than understanding somebody else's.
> There's also the 'not made here' syndrome. So there's a lot of
> refactoring (which it's not bad if it's not refactoring everything at
> once). Having automated test cases for all features would help reducing
> the number of bugs which shouldn't exist. Does the open source developer
> rely more on the many eyeballs than these tests?

None of the things you've pointed out are caused by 100% rewrites.

Cheers,

Matt
msw redhat com
--
Matt Wilson
Manager, Base Operating Systems
Red Hat, Inc.





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