This is a squash commit, the unsquashed development went through many
unstable phases which would break bisects. The unsquashed branch is:
https://github.com/cirosantilli/linux-kernel-module-cheat/tree/v3.0-unsquash
The main improvement of this release was to greatly generalize the testing system.
The key addition was cli_function.py, which allows scripts such as ./run to
be transparently called either from Python or from the command line.
New tests scripts were created using this improved framework: test-baremetal
and test-user-mode.
We were lazy to port some of less important tests to the new setup, TODO's were
added, and we need comes they will be fixed. Getting started is however sacred
as usual and should work.
Other changes include:
- gem5: update to 7fa4c946386e7207ad5859e8ade0bbfc14000d91
- run: --tmux-args implies --tmux
- run: add --userland-args to make userland arguments across QEMU and gem5
Get rid of --userland-before as a consequence.
- bring initrd and initramfs back to life
- build-userland: create --static to make build a bit easier
- gem5: --gem5-worktree also set --gem5-build-id
- remove --gem5, use --emulator gem5 everywhere
Allow passing --emulator multiple times for transparent tests selection
just like --arch.
- test-userland: allow selecting just a few tests
- linux: update to v4.20
- buildroot: update to 2018.08
The main motivation for this was to fix the build for Ubuntu 18.10, which
has glibc 2.28, which broke the 2018.05 build at the m4-host package with:
#error "Please port gnulib fseeko.c to your platform!
- getvar --type input
- failed xen attempt, refactor timer, failed svc attempt, aarch64 use gicv3
- build-doc: exit 1 on error, add to release testing
- build: add --apt option to make things easier on other distros
- build-linux: --no-modules-install
Unsquashed version at v2-rc-unsquashed, but that cannot be merged as it
breaks bisects at several points. All bugs will not bisect to this
humongous change.
It all started with a conversion of the Bash scripts to Python, mainly
because I couldn't stand not being able to properly use --options for
run which has a million options.
Then since that required a full testing, I decided to do all the
refactorings that I had in mind at once, and so I did and it became
v2-rc.
This is the largest patch I have ever done! OMG a few weeks of extra time.
I'm never writing a Bash script for anything that starts getting big again.
Some of the features are:
* separate build-qemu and build-gem5 commands
* common: convert scripts to python. Add --option for everything
* rename build to build-buildroot now that we are splitting all the build
commands, Linux kernel to follow
* move all git submodules to submodules/ and all buildroot packages to
packages/
* refactor the out/ structure. Keep projects on toplevel, because guest
projects separate archs and host ones don't, making a toplevel arch wrong
* do-release: rename to just release
https://stackoverflow.com/questions/16174992/cant-get-argparse-to-read-quoted-string-with-dashes-in-it
* run: add --terminal and explain gem5 pdb
* just track the lvimrc
* store CLI kernel config fragment inside buildlroot to avoid conflicts
* gem5: document m5 initparam
* readme: make a bunch of things awesomer
* readme: fix broken refs
* parsec-benchmark: update to 75d55ac446a43c47efb1044844a108c6c330184c
Could not fetch otherwise.
* gem5: M5_OVERRIDE_PY_SOURCE
As noted however, it is potentially too slow to be useful.
run: unify gem5 and qemu tracing under -T
readme: overhaul tracing documentation from what I've learnt from trace2line
gem5 cache automate benchmarks, and remove results that were scattered all over.
Move console fun under linux kernel section, since not important enough
for getting started.
Set arch x86_64 by default through common on all scripts.
Document that X11 mouse is not moving.
Convert x='' to x= on all scripts, and also fix case ;; indentations.
Add dummy value to QEMU's -trace enable= to prevent warning.
Expand built-in package choice rationale.
Most of it was present inside buildroot/output.* and the rest scattered
on top level.
This came about for the n-th time when we were reviewing QEMU trace file
locations.
On one hand, it would be cool to have per arch traces.
This made buildroot/output.${arch}~/ a natural choice.
But on the other, those traces have nothing to do with Buildroot,
and could potentially interfere with Buildroot build files.
It also feels nicer to have buildroot/ pristine source code only,
and keep all output under a single directory out/
This lead to a re-factoring of count boot instructions, since trace2line
relies on the same tracing mechanism.
Also, now that we have common factored, I've decided to put the traces on
the out dir, to concentrate all outputs in a single place, and allow
easy switching between archs.
