Some general development questions

[Colin Ian King]

On 26/07/16 13:50, Prarit Bhargava wrote:
> 1.  It seems like the correct procedure on this list to review a patch is to
> 
> a) add ACK: to the subject, and
> b) Add an Acked-by: line to the end of the patch.
> 
> Is one notification or the other also acceptable?

The convention is a + b, sometimes we forget one or the other, but
generally we try and do a + b.

We normally require 2 Acks before code is pulled into the git
repository.  If you see code that needs commenting on, feel free to
contribute comments etc.

(Sometimes, when we are very close to release and a trivial patch lands
in the list to fix packaging issues or build issues we may then fall
back to 1 Ack to get things turned around quickly).

> 
> 2.  Is the intent of fwts to test for *specific* known failure cases, or should
> an overall general approach to verification be taken?  For example, I've seen a
> failure in _DSM methods on specific vendor's hardware.  Should I target a test
> for that failure, or make the test as broad as possible?

My opinion is that fwts checks against the specifications, and, if we
know of any special specific issues we can add tests for these too if
they don't go against the specs,

> 
> [This is a loaded question -- in some cases there is a lot of information in the
> tables and it may be better to take a step approach to enhancing the test suite.]
> 
> 3.  debian/ exists.  Is it okay to add a fedora directory that contains a
> buildrpm.sh script & fwts.spec file?

Yep, that's OK if it does not keep on changing too frequently (c.f.
debian changelogs etc) and if you are happy to maintain it.

> 
> 4.  [This one is more of a personal statement and question :)]  I suck at
> Makefiles.  I *really* *really* *really* suck at them.

No problem, we can try and help if you get stuck and need some
incantations to get things sorted for you.

> 
> I have noticed the following issue that I think should be addressed in the
> Makefile.  When executing 'fwts --uefitests' the efi_runtime module is not
> automatically loaded into the kernel.  I think that would be a trivial exercise
> for someone more familiar with the Makefile layout than I am ;) -- who is the
> right person to ping for that request?

So a few things to bear in mind.

1. the efi_runtime module is packaged up in Ubuntu as a DKMS module.
This has special packaging rules to make the dkms package. DKMS is a pain.

2. it is our intention to get this into the upstream kernel to do away
with this complexity. Hopefully, if things go well, this will be in the
4.8 kernel time frame.

> 
> 4a)  Is there any reason that efi_runtime only exists in fwts and not the kernel
> (linux.git) tree?

See above :-)
> 
> P.
> 
Colin