Re: [EXTERNAL] Re: [edk2-devel] generating the test matrix for local CI runs

["Laszlo Ersek" <lersek@redhat.com>]

On 03/04/21 19:32, Bret Barkelew wrote:

> Can you share the specific PR you\x19re failing so that we can talk in
> concrete terms?

Sure.

In January, I posted the following patch series:

  [edk2-devel] [PATCH 0/8]
  ShellPkg, ArmVirtPkg, OvmfPkg: shell usability improvements

  https://edk2.groups.io/g/devel/message/69590
  https://www.redhat.com/archives/edk2-devel-archive/2021-January/msg00070.html
  Message-Id: 20210104154235.31785-1-lersek@redhat.com

This series cleared my local CI just fine (both Windows and Linux VM),
before I posted it.

When review was done, I attempted to merge the series:

  https://github.com/tianocore/edk2/pull/1336

Although the PR does not show it now (or at least I cannot make it show
it), the github CI uncovered a real problem in my patch set, one that my
local run missed.

Namely, I had added an OrderedCollectionLib class dependency to the
(singleton) UefiShellCommandLib instance, but failed to resolve the
OrderedCollectionLib class in the EmulatorPkg and UefiPayloadPkg DSC
files. Consequently, those DSC files would no longer build. My local CI
runs didn't extend to these packages, but -- thankfully -- the
github.com-based CI did.

In v2, I added one extra patch for each of EmulatorPkg and
UefiPayloadPkg DSC:

  [edk2-devel] [PATCH v2 00/10]
  multiple packages: shell usability improvements

  https://edk2.groups.io/g/devel/message/70216
  https://www.redhat.com/archives/edk2-devel-archive/2021-January/msg00700.html
  Message-Id: 20210113085453.10168-1-lersek@redhat.com

(What's more, I tracked down the DSC files in edk2-platforms that used
to build the shell, and I resolved the OrderedCollectionLib class in
them as well -- see the links inside the above-referenced blurbs.)

The v2 CI succeded on github (personal, then merge):

  https://github.com/tianocore/edk2/pull/1337
  https://github.com/tianocore/edk2/pull/1370

In summary, the bug (in my local CI usage) was that I didn't cover
EmulatorPkg and UefiPayloadPkg.

So, I'd like if the local test builds determined the affected set of
packages as well -- I'd like to be *unable* to miss EmulatorPkg and
UefiPayloadPkg in the above type of situation.

Thanks!
Laszlo