Was comparing Option to something that was not an Option

Made `Unit` return value of methods explicit

Report the variable being matched
Improved messages for exceptions caught by
non-exhaustive matches

Remove an incorrect code path in the CIRCT phase that would convert a
"ReplSeqMemAnnotation" into a "-blackbox" argument.  This option has
existed for a long time and this would just cause "firtool" to error.
Users will need to use the dedicated "-repl-seq-mem" options to
"firtool" going forward (which is exactly what they have to do now).

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Modify the ExtModuleImplSpec test to use the MFC and not the SFC.  This
includes one disabled test until firtool 1.30.0 is released.  There is a
silly bug where a supported Annotation (BlackBoxPathAnno) is supported
but rejected during parsing. See:

  - https://github.com/llvm/circt/commit/0285a98d96b8df898e02c5ed9528f869bff80dcf



Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Convert the SelectSpec to use the MFC instead of the SFC.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Convert the AnnotationNoDedup test to use the MFC.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Change the ForceNamesSpec, a test of the forceName Chisel API, to use
the MFC instead of the SFC.  This requires deleting all tests forcing
component names as this is not supported by the MFC.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Deprecate forceName APIs on components.  These are only supported on
modules/instances in the MFC.  (This could be revived in the future if
desired.)

Deprecate the ForceNamesTransform.  This doesn't need to be public given
that the forceName Chisel API is available to users.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Deprecate implementation details related to the loadMemoryFromFile API.
Specifically, deprecate the Chisel-level annotation and the SFC
transform while preserving the Chisel APIs.  The Annotation will, after
Chisel 3.6 be made private and the transform will be deleted.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Previously, they were aliases into experimental but this does not work
very well for ScalaDoc.

Remove loadMemoryFromFile (and implicitly loadMemoryFromFileInline) from
the non-exhaustive list of deprecated things in the Roadmap.  This is
now planned to be supported by MFC [1].

[1]: https://github.com/llvm/circt/pull/4622



Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

* Make unittests that used BasicTester and TesterDriver to use MFC

* Remove compatibility mode tests

* converted LoadMemoryFromFileSpec to circt
made tests ignored for now, pending Adam's effort to make this work

* hierarchy.Utils did not actually use ChiselStage

* Move ChiselEnum spec to MFC

* Refactor how website jobs are run and deployed
* [website] README: remove text about building API docs
* [website] Stop publishing pre-3.5 API Docs, remove links from the website and related Make flow and documentation
Co-authored-by: Jack Koenig <koenig@sifive.com>

* [website] Make Contributors fetching more robust
* [ci] Give website build step access to GITHUB_TOKEN

Fix FIRRTL direction calculations for Vec and OpaqueType

Similarly to Vecs, both the direction of the OpaqueType itself and the
direction of its element can affect the direction of the resulting
firrtl port. Since OpaqueTypes can wrap other OpaqueTypes, this
direction calculation needs to be recursive. The same basic principle
applies to calculating firrtl flip for OpaqueTypes that are elements of
other Records.

For ports of type Vec, the direction of both the Vec itself and the
direction of its elements can affect the direction of the FIRRTL port.
Previously, we did not handle the fact that the Vec's elements could,
themselves, be Vecs, thus making this directionality calculation
recursive. Also, the logic was only applying to regular RawModules and
Modules. The fix now applies to BlackBoxes and ExtModules as well.

* Publish website to gh-pages branch (#2950)
* Update website/README.md

Records now have auto cloneType always enabled.
chisel3.experimental.AutoCloneType is deprecated.

Co-authored-by: Jack Koenig <koenig@sifive.com>

Add FIRRTL Intrinsic support with CIRCT sizeOf intrinsic.

Remove the default option of "-verify-each=false" to firtool.  This
shouldn't be necessary and shouldn't be turned on.  I had left this on
when doing some performance comparisons and it made its way into
upstream chisel.

Signed-off-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

Bring the website (chisel5) in as git subtree

- update README to have proper links to this version of the code
- Move website CI script to top level workflows, give some more unique names
- remove chisel3 submodule dependency from website
- website: Tweak to build within chisel3
- Add firtool as a dependency in CI and the README for the website
- website: remove dependabot file since we have no submodules
- add website to required CI checks

Update .github/workflows/test.yml

Co-authored-by: Schuyler Eldridge <schuyler.eldridge@sifive.com>

git-subtree-dir: website
git-subtree-mainline: b67a1adf
git-subtree-split: 160a018e