Ruby test suite failures with libffi-3.4.2
voxik opened this issue · comments
Hi, libffi-3.4.2 recently landed in Fedora and since that time, I observe strange failures in Ruby test suite:
1) Failure:
TestAutoload#test_autoload_fork [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_autoload.rb:380]:
[ruby-core:86410] [Bug #14634].
Expected #<Test::Unit::AssertionFailedError: Expected #<Process::Status: pid 3249430 SIGABRT (signal 6) (core dumped)> to be success?.> to be nil.
2) Failure:
TestAutoload#test_autoload_fork [/builddir/build/BUILD/ruby-3.1.0/tool/lib/zombie_hunter.rb:6]:
Expected [[3249431, #<Process::Status: pid 3249431 SIGABRT (signal 6) (core dumped)>]] to be empty.
3) Failure:
TestRand#test_fork_shuffle [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_rand.rb:276]:
#<Process::Status: pid 3249809 SIGABRT (signal 6) (core dumped)>.
Expected #<Process::Status: pid 3249809 SIGABRT (signal 6) (core dumped)> to be success?.
4) Failure:
TestRand#test_rand_reseed_on_fork [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_rand.rb:306]:
[ruby-core:41209]
pid 3249817 killed by SIGABRT (signal 6) (core dumped)
5) Failure:
TestIO#test_copy_stream_socket7 [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_io.rb:995]:
Expected #<Process::Status: pid 3251775 SIGABRT (signal 6) (core dumped)> to be success?.
6) Failure:
JSONGeneratorTest#test_broken_bignum [/builddir/build/BUILD/ruby-3.1.0/test/json/json_generator_test.rb:305]:
Failed assertion, no message given.
7) Failure:
TestBeginEndBlock#test_internal_errinfo_at_exit [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_beginendblock.rb:175]:
Expected #<Process::Status: pid 3252235 SIGABRT (signal 6) (core dumped)> to not be signaled?.
8) Failure:
TestProcess#test_signals_work_after_exec_fail [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_process.rb:2428]:
Expected #<Process::Status: pid 3252558 SIGABRT (signal 6) (core dumped)> to be success?.
9) Failure:
TestProcess#test_threading_works_after_exec_fail [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_process.rb:2464]:
Expected #<Process::Status: pid 3252910 SIGABRT (signal 6) (core dumped)> to be success?.
10) Failure:
TestProcess#test_process_detach [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_process.rb:2348]:
Expected #<Process::Status: pid 3253011 SIGABRT (signal 6) (core dumped)> to be success?.
11) Failure:
TestThread#test_blocking_mutex_unlocked_on_fork [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_thread.rb:1223]:
[ruby-core:55102] [Bug #8433].
<false> expected but was
<nil>.
12) Failure:
TestThread#test_fork_in_thread [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_thread.rb:1243]:
[ruby-core:62070] [Bug #9751]
pid 3253221 killed by SIGABRT (signal 6) (core dumped).
Expected #<Process::Status: pid 3253221 SIGABRT (signal 6) (core dumped)> to not be signaled?.
13) Failure:
TestThread#test_fork_while_locked [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_thread.rb:1254]:
[ruby-core:85940] [Bug #14578].
Expected #<Process::Status: pid 3253241 SIGABRT (signal 6) (core dumped)> to be success?.
14) Failure:
TestThread#test_fork_while_locked [/builddir/build/BUILD/ruby-3.1.0/tool/lib/zombie_hunter.rb:6]:
Expected [[3253249, #<Process::Status: pid 3253249 SIGABRT (signal 6) (core dumped)>],
[3253250, #<Process::Status: pid 3253250 SIGABRT (signal 6) (core dumped)>]] to be empty.
Finished tests in 885.270988s, 24.2186 tests/s, 3106.1596 assertions/s.
21440 tests, 2749793 assertions, 14 failures, 0 errors, 56 skips
ruby -v: ruby 3.1.0p0 (2021-12-25 revision fb4df44d16) [x86_64-linux]
make: *** [uncommon.mk:822: yes-test-all] Aborted (core dumped)
Executing the test on itself, they passes just fine, but testing e.g. test/ruby/test_autoload.rb -v -n '/TestAutoload#test_autoload_fork/' together with test/fiddle/test_import.rb makes the reproducer smaller.
Trying to reduce the issue even further, I have reduced the test/ruby/test_import.rb into the following shape:
# coding: US-ASCII
# frozen_string_literal: true
begin
require_relative 'helper'
require 'fiddle/import'
rescue LoadError
end
module Fiddle
module LIBC
extend Importer
dlload LIBC_SO, LIBM_SO
CallCallback = bind("void call_callback(void*, void*)"){ | ptr1, ptr2|
# f = Function.new(ptr1.to_i, [TYPE_VOIDP], TYPE_VOID)
# f.call(ptr2)
}
end
end if defined?(Fiddle)
And use just miniruby:
$ gdb --args ./miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems "./test/runner.rb" --ruby="./miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems" --excludes-dir=./test/excludes --name='!/memory_leak/' test/fiddle/test_import.rb test/ruby/test_autoload.rb -v -n '/TestAutoload#test_autoload_fork/'
GNU gdb (GDB) Fedora 11.1-6.fc36
Copyright (C) 2021 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-redhat-linux-gnu".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<https://www.gnu.org/software/gdb/bugs/>.
Find the GDB manual and other documentation resources online at:
<http://www.gnu.org/software/gdb/documentation/>.
For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from ./miniruby...
warning: File "/builddir/build/BUILD/ruby-3.1.0/.gdbinit" auto-loading has been declined by your `auto-load safe-path' set to "$debugdir:$datadir/auto-load".
To enable execution of this file add
add-auto-load-safe-path /builddir/build/BUILD/ruby-3.1.0/.gdbinit
line to your configuration file "/builddir/.config/gdb/gdbinit".
To completely disable this security protection add
set auto-load safe-path /
line to your configuration file "/builddir/.config/gdb/gdbinit".
For more information about this security protection see the
"Auto-loading safe path" section in the GDB manual. E.g., run from the shell:
info "(gdb)Auto-loading safe path"
(gdb) r
Starting program: /builddir/build/BUILD/ruby-3.1.0/miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems ./test/runner.rb --ruby=./miniruby\ -I./lib\ -I.\ -I.ext/common\ \ ./tool/runruby.rb\ --extout=.ext\ \ --\ --disable-gems --excludes-dir=./test/excludes --name=\!/memory_leak/ test/fiddle/test_import.rb test/ruby/test_autoload.rb -v -n /TestAutoload\#test_autoload_fork/
Download failed: No route to host. Continuing without debug info for /builddir/build/BUILD/ruby-3.1.0/system-supplied DSO at 0x7ffff7fc4000.
Download failed: No route to host. Continuing without debug info for /lib64/libz.so.1.
Download failed: No route to host. Continuing without debug info for /lib64/libgmp.so.10.
Download failed: No route to host. Continuing without debug info for /lib64/libcrypt.so.2.
Download failed: No route to host. Continuing without debug info for /lib64/libm.so.6.
Download failed: No route to host. Continuing without debug info for /lib64/libc.so.6.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
process 13364 is executing new program: /builddir/build/BUILD/ruby-3.1.0/ruby
Missing separate debuginfos, use: dnf debuginfo-install glibc-2.34.9000-36.fc36.x86_64 gmp-6.2.1-1.fc36.x86_64 libxcrypt-4.4.27-1.fc36.x86_64 zlib-1.2.11-30.fc35.x86_64
Download failed: No route to host. Continuing without debug info for /lib64/libz.so.1.
Download failed: No route to host. Continuing without debug info for /lib64/libgmp.so.10.
Download failed: No route to host. Continuing without debug info for /lib64/libcrypt.so.2.
Download failed: No route to host. Continuing without debug info for /lib64/libm.so.6.
Download failed: No route to host. Continuing without debug info for /lib64/libc.so.6.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
Run options:
--seed=54837
"--ruby=./miniruby -I./lib -I. -I.ext/common ./tool/runruby.rb --extout=.ext -- --disable-gems"
--excludes-dir=./test/excludes
--name=!/memory_leak/
-v
-n
/TestAutoload#test_autoload_fork/
# Running tests:
[Detaching after vfork from child process 13401]
[1/0] TestAutoload#test_autoload_fork[New Thread 0x7ffff4ccf640 (LWP 13402)]
[New Thread 0x7ffff4bae640 (LWP 13403)]
[New Thread 0x7ffff4a8d640 (LWP 13404)]
[New Thread 0x7ffff496c640 (LWP 13405)]
[New Thread 0x7ffff484b640 (LWP 13406)]
[New Thread 0x7ffff472a640 (LWP 13407)]
[Detaching after fork from child process 13408]
[Detaching after fork from child process 13409]
[Detaching after fork from child process 13410]
= 0.39 s
1) Failure:
TestAutoload#test_autoload_fork [/builddir/build/BUILD/ruby-3.1.0/test/ruby/test_autoload.rb:380]:
[ruby-core:86410] [Bug #14634].
Expected #<Test::Unit::AssertionFailedError: Expected #<Process::Status: pid 13409 SIGABRT (signal 6) (core dumped)> to be success?.> to be nil.
2) Failure:
TestAutoload#test_autoload_fork [/builddir/build/BUILD/ruby-3.1.0/tool/lib/zombie_hunter.rb:6]:
Expected [[13410, #<Process::Status: pid 13410 SIGABRT (signal 6) (core dumped)>]] to be empty.
Finished tests in 0.392854s, 2.5455 tests/s, 12.7274 assertions/s.
1 tests, 5 assertions, 2 failures, 0 errors, 0 skips
ruby -v: ruby 3.1.0p0 (2021-12-25 revision fb4df44d16) [x86_64-linux]
Thread 1 "ruby" received signal SIGABRT, Aborted.
0x00007ffff78a764c in __pthread_kill_implementation () from /lib64/libc.so.6
Missing separate debuginfos, use: dnf debuginfo-install glibc-2.34.9000-36.fc36.x86_64 gmp-6.2.1-1.fc36.x86_64 libxcrypt-4.4.27-1.fc36.x86_64 zlib-1.2.11-30.fc35.x86_64
(gdb) bt
#0 0x00007ffff78a764c in __pthread_kill_implementation () from /lib64/libc.so.6
#1 0x00007ffff785a656 in raise () from /lib64/libc.so.6
#2 0x00007ffff7844833 in abort () from /lib64/libc.so.6
#3 0x00007ffff4d0a5b8 in dlfree (mem=0x7ffff7bf1010) at ../src/dlmalloc.c:4350
#4 0x00007ffff4d190b1 in dealloc (ptr=0x5555558c1c00) at /builddir/build/BUILD/ruby-3.1.0/ext/fiddle/closure.c:32
#5 0x00007ffff7cb7801 in run_final (zombie=140737300557440, objspace=0x55555555d800) at /builddir/build/BUILD/ruby-3.1.0/gc.c:4011
#6 finalize_list (objspace=objspace@entry=0x55555555d800, zombie=140737300557440) at /builddir/build/BUILD/ruby-3.1.0/gc.c:4030
#7 0x00007ffff7cb80cc in rb_objspace_call_finalizer (objspace=0x55555555d800) at /builddir/build/BUILD/ruby-3.1.0/gc.c:4194
#8 0x00007ffff7ca56eb in rb_ec_finalize (ec=0x55555555dd70) at /builddir/build/BUILD/ruby-3.1.0/eval.c:164
#9 rb_ec_cleanup (ec=ec@entry=0x55555555dd70, ex0=<optimized out>) at /builddir/build/BUILD/ruby-3.1.0/eval.c:256
#10 0x00007ffff7ca5c14 in ruby_run_node (n=0x7ffff7699660) at /builddir/build/BUILD/ruby-3.1.0/eval.c:321
#11 0x000055555555518f in main (argc=<optimized out>, argv=<optimized out>) at ./main.c:47
(gdb) f 3
#3 0x00007ffff4d0a5b8 in dlfree (mem=0x7ffff7bf1010) at ../src/dlmalloc.c:4350
4350 USAGE_ERROR_ACTION(fm, p);
(gdb) l
4345 check_free_chunk(fm, p);
4346 goto postaction;
4347 }
4348 }
4349 erroraction:
4350 USAGE_ERROR_ACTION(fm, p);
4351 postaction:
4352 POSTACTION(fm);
4353 }
4354 }
I have reduced the reproducer:
$ cat fiddle_fork.rb
require 'fiddle/import'
module Fiddle
module LIBC
extend Importer
dlload "libc.so.6", "libm.so.6"
CallCallback = bind("void call_callback(void*, void*)"){ | ptr1, ptr2| }
end
end
error, pid, status = IO.pipe do |r, w|
pid = fork {}
w.close
[r.read, *Process.wait2(pid)]
end
$ RUBYLIB=.:lib:.ext/common:.ext/x86_64-linux:tool/lib LD_LIBRARY_PATH=. ./ruby fiddle_fork.rb
Aborted (core dumped)
I think this might be similar to ffi/ffi/issues/621
I have also reported this issue in Fedora:
I tested this on Ubuntu (with clang and gcc) and macOS with clang using libffi version 3.4.2 but I'm not able to reproduce this. Since the problem started happening after upgrading libffi, could it be a problem with libffi rather than a problem with Fiddle?
We looked at this issue closer, and it seems to be composed of SELinux, FFI Closure, and Forking the Ruby process.
From the linked bug: "In modern libffi 3.4.2 we use memfd_create quite eagerly for SELinux protected systems and such closures are shared between the parent and child process." and "Closures may be inherited by the child, and so the parent and child must coordinate the usage.".
Which would be consistent with our findings.
See the following minimal reproducer that triggers the bug with Fedora 36 and later:
require 'fiddle/closure'
require 'fiddle/struct'
Fiddle::Closure.new(Fiddle::TYPE_VOID, [])
fork { }
GC.startThis describes all required to trigger the bug on the Ruby side. Note that SELinux plays a role in this as well since libffi takes a different approach to allocate the closure on SELinux systems.
The reproducer demonstrates the following:
First, we allocate closure, then we fork the process. The forked process inherits the shared FFI memory. Then that process ends, I think that Ruby frees what was present in that child process which corrupts the memory that is still present in the parent, then we start GC in the parent (to make sure GC runs, in bigger programs this would happen automatically in the background), which results in a SIGABRT with the backtrace noted in the first comment.
The failure we observe in the Ruby test suite follows a very similar pattern to the reproducer when it fails for us in the build. First, some Fiddle tests that allocate Closures run, then some other test forks, and then the GC needs to be triggered. Some tests trigger them manually (e.g. TestHash#test_replace_bug15358), and some are very likely to trigger the GC simply by their memory usage (e.g. TestTrick2015#test_kinaba).
We can see the problem is that this memory usage is not coordinated properly between parent and child process.
On Fedora side we currently disable the closure tests (ruby.spec) to prevent this kind of failure.
The possible mitigations for this are:
- Enable static trampolines in libffi. As noted in the Bugzilla it seems to improve the situation, but there are some blockers we need to resolve first before we use this approach in Fedora.
- Coordinate memory usage between the child and parent processes
- Do not run tests using Fiddle::Closure in the main process that might fork later on.
To the last point, we also found that if the Fiddle::Closure is allocated in the child, it is not a problem for overall status of the test suite, therefore this approach could be used for tests using closure to ensure that the tests can run even with SELinux and whatnot. Additionally, with this approach, one more test would be created with the mentioned reproducer to test the integration of libffi on the current system.
Using the previous reproducer:
require 'fiddle/closure'
require 'fiddle/struct'
fork do
Fiddle::Closure.new(Fiddle::TYPE_VOID, [])
end
fork { }
GC.startThis does not crash the process, unlike the previous reproducer.
I noticed that Ruby tests include assert_separately, which would help with this. I am unsure if it is usable here as it seems unique to the Ruby test suite.
Thx for the patch. I leave the testing to @jackorp thx 😉
Nevertheless, this is fixing the specific use case in the test suite, while this is generic issue. Therefore I wonder if there is a way to e.g.:
- Prohibit the fork if some closure is allocated?
- Warn that this is not right thing to do.
- Make the closure work with fork on Fiddle / ffi level?
- Document this behavior in documentation.
I am not ffi expert, so apologies if I am completely off.
Thanks for the patch!
It improved the situation somewhat but not complemetely. After uncomentting and re-enabling the tests for Ruby and applying the patch I observe new issue in the build logs:
1) Failure:
Fiddle::TestClosure#test_conversion_unsigned_char [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
2) Failure:
Fiddle::TestClosure#test_conversion_unsigned_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
3) Failure:
Fiddle::TestClosure#test_const_string [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
4) Failure:
Fiddle::TestClosure#test_conversion_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
5) Failure:
Fiddle::TestClosure#test_conversion_char [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
6) Failure:
Fiddle::TestClosure#test_conversion_unsigned_long_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
7) Failure:
Fiddle::TestClosure#test_conversion_int [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
8) Failure:
Fiddle::TestClosure#test_conversion_long_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
9) Failure:
Fiddle::TestClosure#test_conversion_unsigned_int [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
10) Failure:
Fiddle::TestClosure#test_conversion_short [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
11) Failure:
Fiddle::TestClosure#test_block_caller [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
12) Failure:
Fiddle::TestClosure#test_conversion_unsigned_short [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
The Ruby test suite failed. Although without a segfault, that is an improvement.
Interestingly there seems to be some other issue with this, as I have another build log:
1) Failure:
Fiddle::TestFunc#test_qsort1 [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_func.rb:87]:
<0> expected but was
<1>.
Here, only a single test failed. Some builds do pass, but there are more of those that do not pass.
I think sometimes there is a (uncollectable maybe?) reference to Fiddle::Closure introduced in the tests. I'll try some local testing and poking around to find what that reference is.
Maybe asserting that there is either 0 or 1 reference in the ObjectSpace would be the way to go, but it requires lengthier testing on my side to confirm that this approach would not lead to a segfault due to the "random" nature of GC.
Thanks for confirming the patch.
It seems that we need a feature that frees a Fiddle::Closure explicitly without GC. e.g.:
closure = Fiddle::Closure.new(...)
closure.free # Free the closure immediately.
Fiddle::Closure.allocate(...) do |closure|
# ...
end # Free the allocated closure when the block is exited.But I don't have a good name for the feature for now... I consider the name in RubyKaigi 2022... (I give a talk today. I consider the name after my talk is finished.)
Hmm, not better. I also applied the previous commit with GC.start to ensure application of the a0ccc6b without modifying the patch too much. I am seeing similar failures as before.
However I have to note, in order to run the whole Ruby test suite together with the patches, I patch the fiddle shipped with Ruby 3.1.2. I may be missing a commit that makes all this works together.
If my patch-only approach fails, I'll try importing fiddle from the newest master branch commit as a gem and see if that improves the situation.
The work is happening on my downstream Fedora Ruby fork branch: https://src.fedoraproject.org/fork/jackorp/rpms/ruby/commits/fiddle_closure_test
Failed build: https://koji.fedoraproject.org/koji/taskinfo?taskID=91998265
relevant test failures, note failure number 14, which is new:
1) Failure:
Fiddle::TestFunc#test_qsort1 [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_func.rb:87]:
<0> expected but was
<1>.
2) Failure:
Fiddle::TestClosure#test_conversion_unsigned_long_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
3) Failure:
Fiddle::TestClosure#test_const_string [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
4) Failure:
Fiddle::TestClosure#test_conversion_unsigned_char [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
5) Failure:
Fiddle::TestClosure#test_conversion_unsigned_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
6) Failure:
Fiddle::TestClosure#test_conversion_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
7) Failure:
Fiddle::TestClosure#test_conversion_char [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
8) Failure:
Fiddle::TestClosure#test_conversion_unsigned_short [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
9) Failure:
Fiddle::TestClosure#test_conversion_int [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
10) Failure:
Fiddle::TestClosure#test_block_caller [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
11) Failure:
Fiddle::TestClosure#test_conversion_short [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
12) Failure:
Fiddle::TestClosure#test_conversion_unsigned_int [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
13) Failure:
Fiddle::TestClosure#test_conversion_long_long [/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:14]:
<0> expected but was
<1>.
14) Error:
Fiddle::TestClosure#test_free:
ArgumentError: wrong number of arguments (given 0, expected 1+)
/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:83:in `block in test_free'
/builddir/build/BUILD/ruby-3.1.2/.ext/common/fiddle/closure.rb:20:in `create'
/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:82:in `test_free'
/builddir/build/BUILD/ruby-3.1.2/tool/test/runner.rb:23:in `<top (required)>'
./test/runner.rb:14:in `require_relative'
./test/runner.rb:14:in `<main>'
Sorry. One more commit: 2530496
Thanks! The state is Much better now.
However this error from the logs is puzzling:
1) Error:
Fiddle::TestClosure#test_free:
NoMethodError: undefined method `assert_false' for #<Fiddle::TestClosure:0x00007f5f73a35fa0 @__name__=:test_free, @__io__=nil, @passed=false, @_assertions=0, @options={:job_status=>nil, :retry=>true, :hide_skip=>true, :repeat_count=>nil, :excludes=>["./test/excludes"], :ruby=>["./miniruby", "-I./lib", "-I.", "-I.ext/common", "./tool/runruby.rb", "--extout=.ext", "--", "--disable-gems"], :filter=>/\A(?=.*)(?!.*(?-mix:(?-mix:memory_leak)|(?-mix:TestAddressResolve#test_socket_getnameinfo_domain_blocking)))/, :verbose=>true, :seed=>10492}, @__gc_disabled__=false, @tracepoint_captured_singlethread=true, @tracepoint_captured_stat=[[#<RubyVM:0x00007f5f7deb3af0>, 0, 0]], @libc=#<Fiddle::Handle:0x00007f5f73a3a2d0>, @libm=#<Fiddle::Handle:0x00007f5f73a3a2a8>>
Did you mean? assert_file
assert_raise
assert_same
assert_raises
assert_all?
assert_false(closure.freed?)
^^^^^^^^^^^^
/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:86:in `block in test_free'
/builddir/build/BUILD/ruby-3.1.2/.ext/common/fiddle/closure.rb:20:in `create'
/builddir/build/BUILD/ruby-3.1.2/test/fiddle/test_closure.rb:85:in `test_free'
I don't understand how it does not know about assert_false... I assume Ruby only has the test-unit available in the tool/lib/test/unit during the tests phase. That one does not have assert_false, however it has refute, which seems to fulfill similar role and is also available in the upstream test-unit.
Alternatively, consider something like assert(closure.freed? == false).
This seems like the single error so far. I'll try out more builds in parallel later, just to be more sure.
For now I am using the following patch for e1221297:
From e1221297fb0177d98c8670e5490c8131227621d5 Mon Sep 17 00:00:00 2001
From: Sutou Kouhei <kou@clear-code.com>
Date: Thu, 15 Sep 2022 06:40:31 +0900
Subject: [PATCH 2/6] test: don't use power-assert
It seems that we can't use it in ruby/ruby.
---
test/fiddle/test_closure.rb | 8 ++------
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/test/fiddle/test_closure.rb b/test/fiddle/test_closure.rb
index 1726db7..13dfa5b 100644
--- a/test/fiddle/test_closure.rb
+++ b/test/fiddle/test_closure.rb
@@ -83,13 +83,9 @@ module Fiddle
def test_free
Closure.create(:int, [:void]) do |closure|
- assert do
- not closure.freed?
- end
+ assert_equal(closure.freed?, false)
closure.free
- assert do
- closure.freed?
- end
+ assert_equal(closure.freed?, true)
closure.free
end
end
--
2.37.3
Note I changed the assert_false to assert_equal(closure.freed?, false).
So far 4 arch specific builds passed with the changes: https://koji.fedoraproject.org/koji/taskinfo?taskID=92035826 I'll run more builds to confirm the results, but this seems successful so far.
Ran 5 builds with 4 different arches each, only problem was unrelated failure on aarch64 on a few of those. Otherwise, nothing fiddle related!
Thanks!
Please use ced671e for assert_false.