Re: Error compiling tests/net/fdpass

[Martin Husemann <martin%duskware.de@localhost>]

On Sat, Sep 01, 2012 at 04:45:51PM +0200, Adam Ciarci?ski wrote:

> I don't quite understand the logic behind this test. Now, two
> programs are being built for every architecture - fdpass64 and
> fdpass32, BUT on sparc64 and amd64 fdpass32 is built as a 32-bit
> binary. On all other architectures (e.g. i386) both fdpass64 and
> fdpass32 are exactly the same, right? Also, what's the point of having
> 32-bit binary on amd64, when the same test can be run on i386, if
> 32-bit testing is required.

The test program does different things when there is both a 32bit and a 64bit
version available, it passes descriptors from a 64 bit process to a 32 bit
process and vice versa.

The way it is done is to simplify set lists, I presume (you'd have to
ask Christos for an authoritive answer).

Obviously some mips variants should do that too, enhancements are welcome.

Martin