Re: Compilation problems with 2.1.20, incorrectly globalized symbols

Keith Owens (kaos@ocs.com.au)
Thu, 16 Jan 1997 16:26:55 +1100


On Wed, 15 Jan 1997 23:37:39 -0500,
Buddha Buck <bmbuck@acsu.buffalo.edu> wrote:
>As far as I know, Linux has refused to dictate coding style -- the
>Documentation/CodingStyle file is a suggestion, not a requirement. To
>quote from that file, "Coding style is very personal, and won't _force_
>my views on anybody...".

There is a difference between a personal coding style and coding
something that adversely affects other modules. If somebody wants to
write a module in Swahili - fine, that's their decision. However if a
module is written so it cannot be linked into the same kernel as other
modules then it is no longer a personal matter. The latter is the
current problem, name space pollution.

>Yes, keeping this from happening again would require work from Linus.
>But I doubt that that is going to happen. As Alan Cox stated, most of
>the people working on this project consider working drivers more
>important than clean drivers, and that is what sets their priorities.

The only work it requires from Linus is a statement that sources should
not pollute the global name space. Once that is stated from on high
then lesser mortals can analyse the kernels, produce reports and niggle
at developers to fix the definitions. OTOH if Linus does not care
about this issue then we are wasting our time trying to fix it.

>This is a major issue, and why these problems are best solved by the
>maintainers of the particular sections.

Agreed, which is why I'm only producing reports and hoping the
developers will look at them. If Linus agrees to a clean name space
then eventually we might produce our own patches but that is a last
resort.