Compatibility issues.

Compatibility issues.

[David Woodhouse]

I would like to have as few occurrences of #if LINUX_VERSION_CODE ... in 
the actual source code of the MTD stuff as possible.

The 2.2.15 kernel contains a new file linux/compatmac.h which provides a 
handful of useful macros which allow you to write portable code.

This is what I've been using for ensuring compatibility between 2.2 and 2.3 
code, and I'd like to provide a similar linux/compatmac.h file to allow us 
to compile for 2.0 as well.

I've added linux/mtd/compatmac.h which currently contains the definitions 
which are missing from Linus' version - but I'd like my file to become 
obsolete as all the required definitions are accepted into the 'official' 
version of compatmac.h in both the 2.0 and 2.2 kernels.

So I've removed all references to the sysdep-2.[12].h files, and moved some 
of the extra definitions into our own mtd/compatmac.h.

If more are required for fixing the 2.0 build, I'd prefer that they be 
added to mtd/compatmac.h, and we can soon delete the sysdep-2.[12].h files 
from the repository.



--
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

compatibility issues

[Tomasz Grobelny]

Recently I switched to the origin/dccp branch of dccp test tree. And noticed 
that dccp communication between kernel build from origin/dccp branch and 
kernel build from origin/master branch is not possible. When kernels are from 
the same branch (no matter if it is origin/dccp or origin/master) the problem 
does not occur. Is it a known issue?
-- 
Regards,
Tomasz Grobelny

compatibility issues

[Tomasz Grobelny]

Recently I switched to the origin/dccp branch of dccp test tree. And noticed 
that dccp communication between kernel build from origin/dccp branch and 
kernel build from origin/master branch is not possible. When kernels are from 
the same branch (no matter if it is origin/dccp or origin/master) the problem 
does not occur. Is it a known issue?
-- 
Regards,
Tomasz Grobelny

Re: compatibility issues

[Gerrit Renker]

| Recently I switched to the origin/dccp branch of dccp test tree. And noticed 
| that dccp communication between kernel build from origin/dccp branch and 
| kernel build from origin/master branch is not possible. When kernels are from 
| the same branch (no matter if it is origin/dccp or origin/master) the problem 
| does not occur. Is it a known issue?
| -- 
The test tree contains fixes for bugs that are in the mainline kernel
and is thus not bug-compatible with the mainline tree (origin/master).

I can't tell from your description where exactly the communication is "not
possible", but one likely cause is the feature negotiation. This is
definitively broken in origin/master: when setting up the receiver with
CCID-2, and the sender with CCID-3, both will end up using these CCIDs with
correspondingly horrible performance; since the origin/master implementation
performs "feature announcement" rather than "feature negotiation".

The test tree has a full implementation of feature negotiation and will
refuse to start communication if parameters are incompatible. It may
allow communication to an origin/master implementation as long as the
origin/master implementation conforms in its behaviour to  RFC 4340,
section 6. Thus the "communication not possible" is not unconditional.

If you think you have found a bug, please provide more detailed information.


The University of Aberdeen is a charity registered in Scotland, No SC013683.