Convert README, INSTALL, and NEWS to Markdown files. #99
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This allows repeatedly polling some command until it succeeds, to some timeout limit.
Sponsored by: The FreeBSD Foundation
`expr ${OPTIND} - 1` can be replaced with $((OPTIND - 1)) to avoid spawning
a new processes. While it's unlikely to make much of a performance
difference for a single execution, it happens once for ever test case so
it will add up for a full testsuite run.
I have been trying to speed up the time it takes to run the testsuite for
CheriBSD on QEMU RISC-V (originally 22 hours). One of the slowest tests is
the pfctl test since ever single invocation of it spends ~90 seconds
generating the list of available tests (even when running only one test).
And this is after the change in r365708 that removes ~200 calls to
atf_get_srcdir(). Looking at truss output it turns out that using atf-sh
results in many tr processes being spawned to perform string substitution.
The _atf_normalize() function tries to remove characters that are not valid
in shell variable names by replacing . and - with _. However, most strings
passed to the function don't contain those characters, so we unnecessarily
fork() a new tr processes (which adds up to a significant slowdown).
With this change we only call tr if the variable contains one of the
illegal characters rather than unconditionally.
Ideally we would do this substitution using only shell builtins, but
unfortunately the string substitution syntax ${var//} is not supported by
POSIX sh and I am not aware of any alternatives.
Basic time measurements on CHERI-QEMU RISC-V usng
`/usr/bin/time /usr/tests/sbin/pfctl_test -l > /dev/null`:
Before:
`90.68 real 89.99 user 0.10 sys`
After:
`14.50 real 14.31 user 0.10 sys`
I.e. 85% of the time was being spent in these tr invocations!
truss -cf before:
```
syscall seconds calls errors
fork 453.330532000 2371 0
cap_enter 0.247944000 633 0
cap_ioctls_limit 0.761099000 1899 0
vfork 0.421971000 2 0
getegid 0.000342000 1 0
getgid 0.000358000 1 0
getuid 0.000324000 1 0
getppid 0.000589000 1 0
geteuid 0.000631000 2 0
getpid 0.000328000 1 0
issetugid 0.450306000 1270 0
write 0.415779000 634 0
__sysctl 0.280936000 634 0
sigprocmask 2.543827000 6340 0
readlink 0.317952000 635 635
read 1.319873000 2532 0
pread 0.327678000 634 0
open 1.399164000 2536 634
munmap 0.301268000 634 0
mprotect 0.566142000 1268 0
mmap 7.117084000 8876 0
fstat 1.326981000 3168 0
dup2 0.954023000 2373 0
close 4.496506000 9979 0
clock_gettime 0.740010000 1902 0
cap_rights_limit 0.770141000 1899 0
cap_fcntls_limit 0.695135000 1899 0
------------- ------- -------
478.786923000 52125 1269
```
After:
```
syscall seconds calls errors
fork 24.675445000 1107 0
cap_enter 0.000340000 1 0
cap_ioctls_limit 0.001009000 3 0
vfork 0.228345000 1 0
getegid 0.000318000 1 0
getgid 0.000323000 1 0
getuid 0.000345000 1 0
getppid 0.000329000 1 0
geteuid 0.000726000 2 0
getpid 0.000335000 1 0
issetugid 0.001554000 4 0
write 0.000539000 1 0
__sysctl 0.000369000 1 0
sigprocmask 0.003684000 10 0
readlink 0.000440000 1 1
read 0.001318000 2 0
pread 0.000556000 1 0
open 0.002250000 4 1
munmap 0.000524000 1 0
mprotect 0.000767000 2 0
mmap 0.011109000 14 0
fstat 0.001743000 4 0
dup2 0.398638000 1108 0
close 1.996684000 4916 0
clock_gettime 0.001047000 3 0
cap_rights_limit 0.001170000 3 0
cap_fcntls_limit 0.000930000 3 0
------------- ------- -------
27.330837000 7197 2
```
atf-sh/atf-check.1[126]: Sentence does not start on new line
Also create a junit xml file.
CirrusCI doesn't like iso-8859-1, hopefully it accepts utf-8. The test outputs should only contain ASCII characters so this shouldn't matter.
Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
There are more of these that need to be modernized, but this will do for now. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
We will be reusing this when we later allow the results file to be closed and reopened for internal test reasons, so do this cleanup in advance. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
Don't bother with different logic based on ctx->resfile, just set ctx->resfilefd up initially with STDOUT_FILENO/STDERR_FILENO and write to the ctx->resfilefd unconditionally later. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
This allows us to set the resultsfile mid-run if we need to, as will be the case for some of our atf internal tests. Notably, some internal tests will fork and do ATF_REQUIRE's that we do -not- want to be written to our outer-test's results file. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
This will be needed for some internal tests that currently fail as they invoke some inner-test that's supposed to fail, but it isn't supposed to write to the outer test's result file. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
All of the tests that fork_and_wait will fork(), fork() again, print some stuff and exit, then the middle-fork parent will wait for the exit and do some checks. These checks are expected to fail in some cases, but what we aren't expecting is for them to use the same resultfile (oops!). Fix this by using our new API to reset the result file to /dev/null in the middle-fork parent, so that our meta-tests are OK. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
If any atf_expect_*() function is called, then we currently close() the resultfile after the expectation is written then proceed. If we end up *failing* that expectation, then we'll venture across this closed fd (unless we're outputting to stdout/stderr) and fail. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
The Cirrus environment may not have CC/CXX set, in which case we should
use some sensible default. This causes problems as the compiled product
actually tries to use ${ATF_BUILD_CC} and ${ATF_BUILD_CXX}, and this fails
miserably as they're ultimately empty.
However, we also don't really need to do any of this if we don't have any
archflags to add, so just don't.
Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
The exception is stdout/stderr, which we won't even attempt in case it's been redirected to a file in a level above us that wants to check expectations or what-have-you. Signed-off-by: Kyle Evans <kevans@FreeBSD.org>
- Unconditionally bootstrap pkg to avoid pkg - Let job name match `uname -r` (13.0-CURRENT) - Chase latest 12.x-STABLE image Sponsored by: The FreeBSD Foundation
Sponsored by: The FreeBSD Foundation
The error is quite obvious and looks like copy/paste mistake.
Single quotes around commands in examples do not help anything, and are actually harmful in case of 'echo foo'.
This adds ATF_{CHECK,REQUIRE}_INTEQ to match the ATF_{CHECK,REQUIRE}_STREQ
macros. This should make debugging failed test cases a lot easier.
Switch the release image from 12.1 (no longer supported by ports) to 13.0 (the latest). Also test on stable/13 and head.
I'm working on a test that will fail due to a kernel bug. However, it can't be executed unless certain hardware is present, and I can't check if the hardware supports the test without running the entire test. So it makes sense to set atf_tc_expect_fail in the beginning and later to atf_tc_skip if the hardware isn't suitable. But ATF doesn't currently allow that.
Quoted multi-word commands will be passed literally to exec rather than split on arguments, which will simply not work. Stop excessive quotes to make sure we don't mislead folks into thinking they're needed, and slap -x on the one example that needs it. This is a follow up to PR freebsd#13.
The previous logic used 2 separate calls to `atf::fs::path::str()` when constructing a `std::vector<char>` to pass to `mkstemp(..)`. This in turn caused grief with how data copying is done in atf-c(3), etc, as the prior code computed the length of the path of an internal buffer in `atf_dynstr` structs. Moreover, the code was manually appending a nul char, which was unnecessary when making the valid assumption that `std::string` is a nul-terminated string. The new code convert the path to an `std::string` once, includes the existing nul char in the buffer, then passes it to mkstemp(3) instead. The code works properly now. Closes: freebsd#76 Signed-off-by: Enji Cooper <ngie@FreeBSD.org>
atf_check: fix std::length_error thrown from temp_file
- atf_utils_cat_file: output `strerror(errno)` to help diagnose why the `open` call failed. - atf_utils_grep_file: output a message with the path name and `strerror(errno)` for parity with `atf_utils_cat_file(..)` and to make diagnosing problems without a corefile possible. Signed-off-by: Enji Cooper <ngie@FreeBSD.org>
This is being done prior to releasing 0.22.
Signed-off-by: Enji Cooper <ngie@FreeBSD.org>
Update the version to 0.23
atf::fs::path has a str() method, and it returns a std::string constructor. I found this odd that a vector is used here to copy std::string contents. As std::string is a typedef of std::basic_string<char>, this additional copy is not needed. Also, remove calling the str() method on error. Signed-off-by: rilysh <nightquick@proton.me>
atf-check.cpp: remove unneeded copying into vector
Improve diagnostics when paths cannot be opened
Revert "atf-check.cpp: remove unneeded copying into vector"
As noted by the reporter, the "descr" metadata property was made optional back in the 0.8 release (af4059b to be specific). Update the documentation posthumously per the code change. Closes: freebsd#63 Reported by: Roland Illig (@rillig) Signed-off-by: Enji Cooper <ngie@FreeBSD.org>
Mark the "descr" metadata property "Optional"
* admin/make-release.sh: automate release creation This does the absolute bare minimum to produce a release. This logic best resembles the expectations for legacy (pre-0.22) releases of ATF. Various files are omitted as they are not required in order to build from a full `autoreconf`'ed source distribution. Signed-off-by: Enji Cooper <ngie@FreeBSD.org>
atf::fs::path has a str() method, and it returns a std::string constructor. I found this odd that a vector is used here to copy std::string contents. As std::string is a typedef of std::basic_string<char>, this additional copy is not needed. Also, remove calling the str() method on error. Notes by ngie@: `buf.data()` returns `const char*` until C++17, so it can't be used in combination with `mkstemp(..)` (it mutates the template argument). Access the internal buffer with the `operator[]` instead. Add the C++17+ form as well along with a TODO comment, so the code can move to std::string::data eventually. Co-authored by: Enji Cooper <ngie@FreeBSD.org> Signed-off-by: rilysh <nightquick@proton.me> Co-authored-by: rilysh <nightquick@proton.me>
ngie-eign
requested changes
Dec 31, 2024
|
Have you run a full build on this change, BTW? Some of the changes you made before resulted in build failures because the file extensions were changed from to |
b5b0b4f to
6ee1800
Compare
|
Formatted with prettier on vscode |
Signed-off-by: Minsoo Choo <minsoochoo0122@proton.me>
Signed-off-by: Minsoo Choo <minsoochoo0122@proton.me>
Signed-off-by: Minsoo Choo <minsoochoo0122@proton.me>
Added a new commit modifying |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Similar to what I did for lutok (freebsd/lutok#11)