There does not seem to be a convenient way to tell git-show-backports to
produce individual patches with numbers. That's what this script does by
calling git-format-patch for each specified commit ID, letting git do all
the painful work (formatting etc). This has been mostly used during
backport sessions but was apparently never committed!
When running with -m (check for missing backports) we often have to fill
lots of information that can be determined automatically the vast majority
of the time:
- restart point (last cherry-picked ID from one of the last commits)
- current branch (HEAD)
- reference branch (the one that contains most of the last commits)
These elements are not that hard to determine, so let's make sure we
can fall back to them when running in missing mode.
The reference branch is guessed by looking at the upstream branch that
most frequently contains some of the last 10 commits. It can be inaccurate
if multiple branches exist with these commits, or when upstream changes
due to a non-LTS branch disappearing in the middle of the series, in which
case passing "-r" will help. But most of the time it works OK. It also gives
precedence to local branches over remote ones for such choices. A test in
2.4 at commit 793a4b520 correctly shows 2.6/master as the upstream despite
2.5 having been used for the early ones of the tag.
For the restart point, we assume that the most recent commit that was
backported serves as a reference (and not the most recently backported
commit). This means that the usual case where an old commit was found
to be missing will not fool the analysis. Commits are inspected from
2 commits before the last tag, and reordered from the parent's tree
to see which one is the last one.
With this, it's sufficient to issue "git-show-backports -q -m" to get
the list of backports from the upstream branch, restarting from the
last backported one.
It happens from time to time while switching between branches and/or
updating after someone else's changes that regtests are run by accident
on the wrong binary, typically the one the tests were run on during
development and not with the latest adaptations. And obviously it's
when this happens that we break the CI. There are various causes to
this but they all come down to humans context-switching a lot, and
there's no real fix for this that doesn't add even more burden hence
increases the overhead. However we can help the human detect such
mistakes very easily.
This change here will compare the version of the haproxy binary to
the version of the tree, and will emit a warning in the regtest output
if they do not match, regardless of the outcome of the test. This is
sufficient in case of failures because these are quickly glanced over,
and is sufficient as well in case of accidental success because the
warning is the last message. E.g:
########################## Starting vtest ##########################
Testing with haproxy version: 2.7-dev10-cfcdbc-38
Warning: version does not match the current tree (2.7-dev10-111c78-39)
0 tests failed, 0 tests skipped, 182 tests passed
This should not affect builds made out of a git tree because the version
is retrieved using "make version", or exactly the same way as it's passd
to the haproxy binary. We just need to know what "make" command to run,
so $MAKE is used primarily, falling back to "make" then to "gmake". In
case all of these fail, we just ignore the version check. This should be
sufficient to catch human mistakes without affecting the CI.
Since Marko announced at HAProxyConf 2022 that the data plane API is
mostly complete and will now follow the same release cycle as haproxy
starting with 2.7, it's probably the right moment to encourage users
to start trying it so that we can hope to migrate all the painful
discovery stuff there in a not too distant future.
Let's just point to the latest release for now. We'll see in the future
if we need to adapt the link depending on the branch.
Some components like Discourse were already redirecting to https. Other
ones like docs and git are covered by the certificate, and finally
switching the advertised scheme for www should increase the ratio of
H2 and H3 in the stats (resp 8.9 and 1.9%) and possibly help spot new
issues.
Now when publishing a release, if make-releases-json is usable at the same
place, it will be called to regenerate the json index of available versions.
This will be used to rebuild a releases.json file in each download
directory. It only relies on existing files and sorts them by version,
appends known signatures (md5/sha256) and marks the most recent one as
the latest release.
This aims at addressing github issue #1537.
The list of URLs now also adds pending bugs, reviewed bugs, and code
reports. The redirect is performed on haproxy.org since github URLs
are far too large here.
For each test, the version number is evaluated using a call to awk,
which can be slow to start depending on the versions and OS. This is
only needed for a printf() call to keep only leading digits of each
component, multiply them by 1000 and pad them to 3 digits, something
that's clearly doable in plain shell in a portable way. This is what
this patch does, and it saves yet another 400 ms here on the full
test sequence.
grep is used in the arguments loops to check for features such as OPENSSL
or LUA or services like prometheus-exporter. Let's just look for the words
inside the list, which requires to prepend a delimitor at the beginning of
the list and add one at the end.
run-tegtests is starting to take a lot of time to spot which tests are
eligible, because for each test file a lot of "sed" sub-processes are
launched. This commit eliminates calls to sed by using the shell's
internal processing and parsing the VTC file only once. Instead of
extracting each option one by one from the file, all entries that look
like a valid option are passed to a single case/esac statement and their
value is extracted. Splitting into lists is simply done by adjusting the
IFS depending on the list's delimiter, which, contrary to the // pattern
modifier, is supported on every shell.
This was tested on both bash and dash, and the tests' execution time
dropped by 31% from 8.7 seconds to 6.0 seconds.
The early version of the script used to support passing non-branch
arguments but as it evolved we lost that option. Let's use "--" as a
delimiter after the branch(es) to pass optional file names to filter
on. This is convenient to list missing patches on a specific set of
files.
I forgot about OpenSSL 1.0.2, which neither supports the build_sw target
to build only the software, nor reliably supports parallel builds. Given
that we're building 1.0.2 and 3.0.0, let's stay on the safe side and
keep 1.x sequential.
1/4 of the OpenSSL build time is spent building the docs, let's just
build the software and not the doc, by replacing the "all" target
with "build_sw". With this my build time drops from 1'28 to 1'09.
Nothing was done for the other libs, as it's unknown whether they
provide specific build targets.
Running the "make all" phase on my machine with -j$(nproc) shrinks the
build time from 4'52 to 1'28. It will not be that big of a change in
the CI since it looks like two CPUs are exposed, but it should still
remain a net win. Let's enable it. The install phase obviously remains
sequential however.
In patch 3bad3d5 ("BUILD: Makefile: exclude broken tests by default"),
the default setting of the REGTESTST_TYPE variable was set in the
Makefile instead of the run-regtests.sh script.
Doing it in the Makefile was breaking the use of this environment
varible with make ( REGTESTS_TYPES=slow,default make reg-tests )
This patch move the default setting from the Makefile to
run-regtests.sh. It also change the documentation in `make
reg-tests-help` about the default value.
This patch should be backported where 3bad3d5 is backported.
I badly tested my previous patch forgetting to remove the "+" testing
present in options, and not in services; the list of services do not
have any "+" at the beginning of each service
this patch is fixing commit aabde71332 ("MINOR:
reg-tests: add a way to add service dependency")
Signed-off-by: William Dauchy <wdauchy@gmail.com>
I was looking at writing a simple first test for prometheus but I
realised there is no proper way to exclude it if haproxy was not built
with prometheus plugin.
Today we have `REQUIRE_OPTIONS` in reg-tests which is based on `Feature
list` from `haproxy -vv`. Those options are coming from the Makefile
itself.
A plugin is build this way:
EXTRA_OBJS="contrib/prometheus-exporter/service-prometheus.o"
It does register service actions through `service_keywords_register`.
Those are listed through `list_services` in `haproxy -vv`.
To facilitate parsing, I slightly changed the output to a single line
and integrate it in regtests shell script so that we can now specify a
dependency while writing a reg-test for prometheus, e.g:
#REQUIRE_SERVICE=prometheus-exporter
#REQUIRE_SERVICES=prometheus-exporter,foo
There might be other ways to handle this, but that's the cleanest I
found; I understand people might be concerned by this output change in
`haproxy -vv` which goes from:
Available services :
foo
bar
to:
Available services : foo bar
Signed-off-by: William Dauchy <wdauchy@gmail.com>
It takes so much time to write an announce message that sometimes it's
annoying not being able to start the work while a fix is being finished.
With the new "-p" argument, announce-release will allow to prepare the
announce message for the current HEAD and with no tag yet. It will
restart from the last tag and automatically increment the version using
the same algorithm as create-release so that everything is accurate. It
should then be easier at the last moment to just include the final entry
by hand when the last fix finally arrives. For convenience, this argument
also allows to create an announce from another branch than master.
By having three variables it will be easier to preset the version and
the tag separately. One contains the announced version, another one the
associated tag and the last one the final commit ID (used as the ending
point before the release). This initially allows to check for the HEAD
matching the tag only when the version was not forced, hence re-announce
already tagged versions after some extra commits were added for example.
Enabling memory poisonning is often pretty effective for detecting
uninitialized structure fields. Let's enable it by default and let
the user change the arguments at will (e.g. forcing some memory limits
or disabling a poller). This will work with the latest vtest version
to date (02a9bc1).
after 73b520b958 variables SSL_LIB, SSL_INC
are not set, but still used by BoringSSL builds. That leads to error
(I wish we could stop on such errors) and using stock openssl instead
of boringssl
It's cumbersome to copy-paste a commit ID into another window after having
typed "git cherry-pick -sx", so let's have the suggested output format of
git-show prepare this line just before the subject line, it remains at a
stable position on the terminal when searching for "/^commit". One just
has to copy-paste the line into another terminal will result in the commit
being properly picked.
We've never used the output of the rightmost branch with this tool,
and it systematically causes two identical outputs making the job
harder during backport sessions. Let's simply remove the right part
when it's identical to the left one. This also adds a few line feeds
to make the output more readable.
