Activity
From 08/12/2021 to 08/18/2021
08/18/2021
-
11:04 PM Feature #18083: Capture error in ensure block.
- There are lots of ways to achieve this but the reality is there is a lot of code which uses `$!` in the ensure block which is incorrect for the reasons already explained.
Your proposed ideas are fine, except that RuboCop will warn you a... -
05:09 PM Feature #18083: Capture error in ensure block.
- If you only need to know if there is an `Exception` this seems easier:
Using the example from https://bugs.ruby-lang.org/issues/15567#note-27
```ruby
begin
...
rescue Exception => exception
transaction.abort
raise exception
ensure
... -
- The second snippet seems clear, the first one much less.
In what cases `error` is set? If it's a StandardError? If it's an Exception? If it's any kind of "Ruby exception"?
IMHO this pattern is rarely enough needed that the second sni... -
- By the way, perhaps we should consider deprecating `$!` since its usage can lead to incorrect behaviour.
-
- As discussed in https://bugs.ruby-lang.org/issues/15567 there are some tricky edge cases.
As a general model, something like the following would be incredibly useful:
``` ruby
begin
...
ensure => error
pp "error occurred" if... -
09:29 PM Bug #17052: Ruby with LTO enabled has issues with SIGSEGV handler
- vo.x (Vit Ondruch) wrote in #note-22:
> The dumps should be available here: https://vondruch.fedorapeople.org/objdump.tar.gz
I'm afraid that this contains only `.debug_info` and not `.debug_abbrev`.
It seems that `objdump` can dum... -
08:19 PM Bug #17052: Ruby with LTO enabled has issues with SIGSEGV handler
- xtkoba (Tee KOBAYASHI) wrote in #note-21:
> Isn't it possible that either section is broken?
Well, it should be DWARF5, which is pretty recent and on ppc64le, so there might be some surprises ...
> ...
The dumps should be availabl... -
- The message
```
1267: Abbrev Number 106 not found
```
means that `addr2line.c` could not find an abbrev table entry in `.debug_abbrev` referenced from `.debug_info`. Isn't it possible that either section is broken?
It might be... -
- The result with the latest patch is significantly different:
~~~
$ gdb --args ./miniruby -e'Process.kill("SIGSEGV",$$)'
GNU gdb (GDB) Fedora 10.2-6.fc35
Copyright (C) 2021 Free Software Foundation, Inc.
License GPLv3+: GNU GPL ver... -
- An edited patch is attached so that it applies cleanly to the current master head.
-
08:44 PM Bug #18117 (Closed): Segmentation fault when yielding values from Ractors during GC sweeping
- [Link for GitHub PR containing the fix](https://github.com/ruby/ruby/pull/4755)
Ractors may invoke `rb_objspace_reachable_objects_from` when yielding values back to the main-Ractor. If this occurs during a sweeping pass of the GC, the... -
- Why does IO#internal_encoding (and external_encoding) raise if the IO is closed?
They could just return the encoding, it's still stored in in the IO instance, and we are not trying to access the `fd`, right?
JRuby 9.2.17.0 does not r... -
03:14 PM Bug #18077: Marshal.dump(closed_io) raises IOError instead of TypeError
- https://github.com/ruby/ruby/pull/4749 is another fix without the above side effect. Is it OK?
-
02:25 PM Feature #12075: some container#nonempty?
- p8 (Petrik de Heus) wrote in #note-25:
> What about `ary.filled?` ?
> ...
I think `filled?` is a perfect method name. You nailed it.
Or, if you are going with past participles, perhaps `loaded?` may also work. -
01:21 PM Feature #12075: some container#nonempty?
- Boolean `size?` looks good to me. I vote for it.
My first choice was `nonempty?` because this is very common word in popular algorithm text books, e.g., Knuth, Cormen, Sedgewick, Tarjan, Aho, Hopcroft, etc.
However some people pointed ... -
12:37 PM Revision 0b7969b6 (git): Silence LoadError only if it is for `rubygems` itself
- Co-authored-by: David Rodríguez <deivid.rodriguez@riseup.net>
Co-authored-by: Hiroshi SHIBATA <hsbt@ruby-lang.org> -
09:24 AM Revision d668cd18 (git): rb_fix2uint should use FIXNUM_NEGATIVE_P
- rb_num_negative_int_p is equivalent to calling the "<" method on
Integer (and checking whether it is overridden), where in this case we
are interested in whether the "actual" value can fit inside an unsigned
int.
This also was slow beca... -
09:18 AM Bug #18084 (Closed): `JSON.dump` can crash VM.
- Applied in changeset commit:git|574f3af36ea008896419d2e2407f8fce83bdcd2e.
----------
Rewind execution tags more at stack overflow [Bug #18084] -
- https://github.com/ruby/ruby/pull/4751
-
- @nobu suggested the following code could be updated:
~~~ clang
sigill(int sig SIGINFO_ARG)
{
check_reserved_signal("ILL");
#if defined __APPLE__
CHECK_STACK_OVERFLOW();
#endif
~~~
we could also do this on Linux? -
- Minor improvement to formatting.
-
- ~~~ ruby
require 'json'
x = {}; x[:x] = x
JSON.dump(x) # => nil
JSON.dump(x) # => SIGILL
~~~
Results from our friendly all ruby bot:
~~~
:x: 1.8
-:1:in `require': no such file to load -- json (LoadError)
from -:1
e... -
-
-
- https://bugs.ruby-lang.org/issues/18084 maybe regression.
-
- - [Feature #18083] Capture error in ensure block.
- Is it acceptable? -
- I will make a more limited proposal.
08/17/2021
-
10:55 PM Bug #16906: Calling Thread#thread_variable? in IRB sometimes produce wrong result
- This needs to be backported to 2.7.x. It wasn't fixed in 2.7.4.
-
- Proposed patch:
```diff
--- a/ext/digest/md5/md5.c
+++ b/ext/digest/md5/md5.c
@@ -225,7 +225,7 @@
uint32_t xbuf[16];
const uint32_t *X;
- if (!((data - (const uint8_t *)0) & 3)) {
+ if (!(((uintptr_t)data) & ... -
05:25 PM Revision 58bd9434 (git): Replace f_boolcast with RBOOL macro
- * Move f_boolcast definination
* Remove f_boolcast macro defination
* to -
04:26 PM Misc #18082 (Open): FileUtils.remove_entry_secure has inconsistent document
- `FileUtils.remove_entry_secure` and `FileUtils.rm_r` have inconsistency about the vulnerability condition the in the documentations.
`remove_entry_secure` document:
https://github.com/ruby/ruby/blob/6a9bfa4d9387b9d8f07f43f4546437be57... - 03:00 PM Revision 6a9bfa4d (git): * 2021-08-18 [ci skip]
-
- FWIW, another bug is found in `addr2line.c` (#18081), which I believe is irrevant to the issue here, though.
-
- A revised patch (from #note-17) is attached so that `abbrev_number >= 256` is handled properly, although it might be practically not going to happen.
-
- It turns out that `di_read_debug_abbrev_cu` in `addr2line.c` does not work correctly when `abbrev_number` is not aligned in ascending order for each CU entry of `.debug_abbrev`. A patch is attached to fix this.
Note again that I canno... -
- When both `.debug_rnglists` and `.debug_ranges` sections exist, `DW_AT_ranges` can refer to a wrong section, potentially causing a segmentation fault.
`DW_AT_ranges` should refer to `.debug_rnglists` for DWARF version 5, and `.debug_r... -
- https://github.com/ruby/date/commit/de7dca353f
- 12:44 PM Revision 48145282 (git): [DOC] Fix the rdoc for File::Stat#size? [ci skip]
-
- And the NULL string must be of size 0.
-
- Eregon (Benoit Daloze) wrote in #note-7:
> I'm confused, why does `Class#allocate` not simply use the alloc function set with `rb_define_alloc_func`?
You're confused by `Class#allocate` which calls the allocator set for each `T_DATA`... -
- I'm confused, why does `Class#allocate` not simply use the alloc function set with `rb_define_alloc_func`?
Then there would be no need to undefine/redefine allocate, isn't it?
(IMHO users should never call `allocate` directly because... -
- mame (Yusuke Endoh) wrote in #note-5:
> How about raising an exception when attempting to allocate a `T_DATA` object with the default allocator?
Though I agreed it once, no way to tell if the given class is `T_DATA` in the default al... -
06:01 AM Bug #18007: Help developers of C extensions meet requirements in "doc/extension.rdoc"
- How about raising an exception when attempting to allocate a `T_DATA` object with the default allocator?
-
- Actually thread-local might not be enough, we'd probably want Fiber local, i.e., events which are on the same execution stack and contain that enable block call on the stack.
Is there any way to have a TracePoint for a particular Fiber? -
- mame (Yusuke Endoh) wrote in #note-7:
> Note that the second `trace.enable` does not print `#<TracePoint:line t.rb:4 in 'foo'>`.
A very good point, that I thought about when talking a bit with Koichi about this, some time before fili... -
05:24 AM Bug #16889: TracePoint.enable { ... } also activates the TracePoint for other threads, even outside the block
- mame (Yusuke Endoh) wrote in #note-7:
> `TracePoint#enable(target: :block)` does not follow events in a called method. Is this really intentional?
> ...
This matches the behavior of passing the block explicitly and using it as the targ... -
05:19 AM Bug #16889: TracePoint.enable { ... } also activates the TracePoint for other threads, even outside the block
- +1 for Jeremy #6.
I also agree `TP#enable{ ... }` should be thread-local without any options, but I don't think it is valuable to change the default behaver (break compatibility). -
- `TracePoint#enable(target: :block)` does not follow events in a called method. Is this really intentional?
```
trace = TracePoint.new(:line) {|tp| p tp }
def foo
1
end
trace.enable do
foo
end
#=>
# #<TracePoint:line t... -
08:15 AM Feature #17750 (Closed): Update Unicode data to Unicode Version 13.0.0
- Closed (finally) with revision 21fd83a823.
-
- Mention the update to Unicode Version 13.0.0 and Unicode Emoji
Version 13.1 in NEWS.md. This completes issue #17750. [ci skip] -
- Resolved with commit fd7f61c.
It turned out the problem was that in emoji-variation-sequences.txt, lines look like
`0023 FE0E ; text style ; # (1.1) NUMBER SIGN`
whereas in the other two files that we processed the same... -
- duerst (Martin Dürst) wrote:
> In particular, I'm a `puts` debugger, but `puts` debugging doesn't seem to work well when testing.
I found that adding and using the following method in the Test... class allowed me to use `log_test` ... - 07:30 AM Revision de764ae8 (git): * 2021-08-17 [ci skip]
-
- Thx for looking into this. I just wanted to point out that this has low importance for me, since the default route was added in mock: https://github.com/rpm-software-management/mock/commit/a1a3ddaa41c0a2b5d17b004bdf3b87a1337eaecc
-
- The emoji data in emoji-variation-sequences.txt was not used for
in test/ruby/enc/test_emoji_breaks.rb, for unknown reasons.
It turned out that the format of each of the emoji data/test files
is slightly different, and that we didn't tak... -
- One line pattern matching with a method return value with parameters which are not surrounded by parenthesis raises syntax error.
I think it is not intentional, but nobu said it's hard to support because of parse.y limitation.
```r... -
05:43 AM Feature #12075: some container#nonempty?
- What about `ary.filled?` ?
It nicely pairs with empty and is not too long.
It’s also a property of the container. And unlike `full?`, it doesn’t imply there is no more room. -
02:31 AM Feature #12075: some container#nonempty?
- I tend to like `ary.size > 0` more than `!ary.empty?` because the former literally has a "positive" nuance and therefore it's more readable in many cases. In that sense, the name `nonempty?` does not sound ideal to me.
So, I came up wi... -
- ```
building spec/ruby/optional/capi/ext/array_spec.so
lld-link: error: x64-ucrt-ruby310.dll: bad file type. Did you specify a DLL instead of an import library?
clang-13: error: linker command failed with exit code 1 (use -v to see in...
08/16/2021
-
- Ah, I can see there is missing the `read_abstract_origin` on the backtrace.
-
- This is the GDB output:
~~~
(gdb) r
Starting program: /builddir/build/BUILD/ruby-3.0.2/miniruby -e Process.kill\ :SEGV,\ \$\$
Download failed: No route to host. Continuing without debug info for /lib64/libz.so.1.
Download failed:... -
- Aren't there any subtle differences in GDB outputs? If fixed as intended, the value of `abstract_origin` can't get such large as 760312297580551 (0x2b380006a6807).
-
- Thank you for looking into this.
Unfortunately, I have tried commit:git|72317b333b85eed483ad00bcd4f40944019a7c13 and I see no difference.
BTW I test Ruby 3.0.2 including the following commits:
1) commit:git|9e5105ca451d6d38eb2d0... -
- `rb_fix2uint` checks if the FIXNUM being converted overflows the `unsigned int` type. Previously this check was being done with `rb_num_negative_int_p`, which is equivalent to calling the `<` method on Integer (and checking whether it is...
-
07:35 PM Misc #18039: DevelopersMeeting20210819Japan
- * [Feature #18007] C extension "allocation function" check
* Should we add a config option and put in 3.1, or just wait for 3.2? -
- One concern for that [PR], probably may not be a matter, is closed `IO`s will no longer raise `IOError` on other `Encoding` operations too.
[PR]: https://github.com/ruby/ruby/pull/4746
```
$ ./ruby -v -e 'p Encoding.compatible?(Enco... -
- Marshal.dump is expected to raise a `TypeError` for unmarshallable objects. But closed streams raise an `IOError`:
```
$ ruby -e "io=IO.pipe.first; io.close; Marshal.dump(io)"
-e:1:in `internal_encoding': closed stream (IOError)
f... -
- Just FTR, Clang/LLVM `__builtin_longjmp` is broken for Sparc targets: https://bugs.llvm.org/show_bug.cgi?id=51489
-
- ```
compiling ../../../../ext/digest/md5/md5.c
../../../../ext/digest/md5/md5.c:228:17: warning: performing pointer subtraction with a null pointer has undefined behavior [-Wnull-pointer-subtraction]
if (!((data - (const uint8_t *... -
-
-
-
- On emscripten `intptr_t`, `uintptr_t`, `ptrdiff_t` and so on are
defined as `long`, but `PRIdPTR` and so on defined as `int`. -
- Select the C source if exists, otherwise the assembler source,
instead of selecting by the coroutine type. - 12:07 AM Revision 4cc44bd8 (git): * 2021-08-16 [ci skip]
- 12:06 AM Revision f8f6d989 (git): coroutine use asm version for handful of architectures for freebsd.
08/15/2021
-
-
11:55 AM Misc #18039: DevelopersMeeting20210819Japan
- * [Feature #16182] Should `expr in a, b, c` be allowed or not? (ktsj)
* How about allowing brackets/braces to be omitted in one-line pattern matching? -
- How about allowing brackets/braces to be omitted in one-line pattern matching?
Now that we use `=>` in one-line pattern matching, that syntax can't be made into an argument whether we allow omission or not. -
- Applied in changeset commit:git|2aa6826e81ec880f5e98fa995e44fae0819be95d.
----------
Extract the wrapped value when yydebug [Bug #18075] -
-
- [0] => [0, *, a]
#=> [0] length mismatch (given 1, expected 2+) (NoMatchingPatternError)
Ignore test failures of typeprof caused by this change for now.
08/14/2021
-
07:36 PM Bug #18000: have_library doesn't work when ruby is compiled with --disable-shared --disable-install-static-library
- No objection from me to drop support for the `--disable-shared
--disable-install-static-library` config. I agree with Jean that it's an
improvement to give a clear message and fail fast.
> With that combination, shouldn't we drop the he... -
- byroot (Jean Boussier) wrote in #note-6:
> But yeah I wish I could `--disable-shared` without having to ship a huge 100MiB+ static library, but since I don't quite understand what it would take to support this combination of flags, I ha... -
12:27 AM Bug #18000: have_library doesn't work when ruby is compiled with --disable-shared --disable-install-static-library
- > Not sure how byroot (Jean Boussier) or alanwu (Alan Wu) will feel about this approach, though.
At the very least I'd say it's progress as my attempt to ship ruby builds with this combinations of flags would have immediately failed d... -
- Applied in changeset commit:git|147bdcc436c888a56f81e190d192cd9312015836.
----------
[Win32] put a space before configure options [Bug #17588]
Not to be concatenated with the preceding `--with-`* option. -
- Not to be concatenated with the preceding `--with-`* option.
- 03:27 PM Revision 717f0c56 (git): * 2021-08-15 [ci skip]
-
-
- Applied in changeset commit:git|72317b333b85eed483ad00bcd4f40944019a7c13.
----------
Ignore `DW_FORM_ref_addr` [Bug #17052]
Ignore `DW_FORM_ref_addr` form and other forms that are not supposed
to be used currently. -
- Ignore `DW_FORM_ref_addr` form and other forms that are not supposed
to be used currently. -
-
- Get rid of tons of linker warnings that it could not find object
file symbol for every symbols, when targeting darwin. -
-
-
-
-
-
01:13 PM Revision 9f88135f (git): [ruby/readline-ext] Give up Editline
- https://github.com/ruby/readline-ext/commit/d924cfc1e6
-
- https://github.com/ruby/reline/commit/8b55fc8fee
-
- https://github.com/ruby/reline/commit/fa54e521da
-
- There used to be a process that did not want to be interrupted by
SIGINT, so it was trapped, but that process is no longer there.
https://github.com/ruby/reline/commit/ba7252a5db -
-
09:35 AM Bug #16490: mkmf.rbの非互換な変更
- 本件は少なくともRuby 2.7.3辺りでは起きなくなっていますので、チケット自体はcloseしていただいても問題ないと考えています。
> dependファイルを直してコンパイルしてみてもエラーになりますので、
これは、コンパイラのオプションに-Werror付けている環境の問題と考えられます。
実際、-Werrorの付かない環境ではウォーニングを出すものの、正常にコンパイルは完了しています。
もちろん、ウォーニングを出さない様に修正すべきでしょうけれ... -
- `depend`ファイルを直してコンパイルしてみてもエラーになりますので、tmail gem自体をもっと修正する必要があると思います。
```
tmailscanner.c:105:11: error: implicit declaration of function 'rb_get_kcode' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
tmp = rb_get_kcode... -
09:10 AM Bug #16836: configure-time LDFLAGS leak into ruby pkg-config file
- Thanks for taking a look!
jeremyevans0 (Jeremy Evans) wrote in #note-1:
> According to the history, adding `DLDFLAGS` to `Libs` was a deliberate decision made back in 2010 in commit:51d25ca8c0eb7da192f5bdf2729fc856e8f81a9d. Assignin... -
- I will check it.
-
08:06 AM Bug #18048: Thread#join can break with fiber scheduler unblock fails or blocks.
- I create the backport patch including 050a89543952a2c9e7c9bc938f4fdb538f6c9278 and 13f8521c630a15c87398dee0763e95f59c032a94 and push to my branch. See https://github.com/ruby/ruby/pull/4686/files.
But on the branch, make btest hangs o... -
- Recently I read similar topic again elsewhere. They pointed
* in most cases we have something to do on each duplicate element if any duplicate detected, e.g., reporting all duplicate elements as an error message
* `uniq?` looks slightly... -
04:20 AM Bug #18075 (Closed): Crasher using ripper + yydebug
- `method `inspect' called on unexpected T_NODE object` via the following with `-d` (see output at end):
```
#!/usr/bin/env ruby -wvs
$d ||= false
require "ripper/sexp"
src = "'woot'"
rip = Ripper::SexpBuilderPP.new src
ri... -
08/13/2021
-
-
- Reverting the line removed from mkmf.rb in r67033 (commit:3d1c86a26f0c96a9c1d0247b968aa96e6f3c30bb) allows for the tmail gem to install on the master branch. I've submitted a pull request for that: https://github.com/ruby/ruby/pull/4738
-
- According to the history, adding `DLDFLAGS` to `Libs` was a deliberate decision made back in 2010 in commit:51d25ca8c0eb7da192f5bdf2729fc856e8f81a9d. Assigning to @nobu as he was the one that made the change. I'm guessing there are env...
-
- I reviewed the mruby `presym.rake` file and was able to come up with a self contained reproducible example, which I have attached. I confirmed the issue is still present in Ruby 3.0 (this error shows the issue happened between 600 and 1...
-
- nobu (Nobuyoshi Nakada) wrote in #note-4:
> We should disallow that combination, I think.
> ...
Here's a pull request to disallow the combination of `--disable-shared --disable-install-static-library`: https://github.com/ruby/ruby/pull... -
- We should disallow that combination, I think.
Otherwise, should drop also the headers. -
05:26 PM Revision 963f2e01 (git): Reduce the amount of false positive notifications
-
-
- This reverts commit b64f81c81729bbc248d19af01cafde88eb60fdc7.
It seems to cause a problem in --jit / --jit-wait CIs. Reverting for now. - 04:10 PM Revision 15c9c784 (git): * 2021-08-14 [ci skip]
-
-
02:24 PM Bug #17588: Cannot build with library path with spaces on MSVC
- I think there's an issue somewhere in the configuration for msvc, missing a space between two conf args it seems.
# Try 1
```
..\ruby\win32\configure.bat --disable-install-doc --enable-bundled-libffi --with-opt-dir=C:/Users/julien/Sof... -
- I'll try again, I'll report back. I see how your change could have fixed the enc/lib specifically indeed.
-
- I tried today with master (ac4d53bd461ff386cd45fdd484ffb6b628a251ad). So it isn't fixed by https://bugs.ruby-lang.org/projects/ruby-master/repository/git/revisions/8a5e161cd9a251054e851d46ccc367609a31c1c7
-
- jmarrec (Julien Marrec) wrote in #note-4:
> My first thing is I've noticed that if I strap --with-static-linked-ext to the configure options, then no matter what I try, I always have to build ruby twice. There's a dependency issue somew... -
- I am revisiting this with the latest master today.
My command has quotes, and I am not changing the `--with-openssl-dir="C:/Program Files/OpenSSL-Win64"` which works plenty fine without `--static-linked-ext`
https://github.com/jmar... -
- It turns out that `DW_FORM_ref_addr` is not interpreted correctly. A patch is attached to just ignore that form (and other forms that are not supposed to be used currently in `addr2line.c`). This ignorance does not seem to affect the qua...
-
- Now I can reproduce this on my qemu-user-emulated powerpc64le-linux-musl environment:
```
$ powerpc64le-musl-qemu -strace ./miniruby -e 'Process.kill :SEGV, $$'
(...)
24209 readlink("/proc/self/exe",0x0000004000490bd8,4096) = 75
2... -
- FYI, I changed the MJIT implementation in commit:ac4d53bd461ff386cd45fdd484ffb6b628a251ad to not disable JIT-ed code when a TracePoint for class events gets enabled. So, as long as Zeitwerk only uses TracePoints for class events, we no l...
-
- events get enabled
-
-
-
-
- Applied in changeset commit:git|573eef7c897ff938539270e605445da13093d664.
----------
Indicate the slow path of `Check_Type` never return [Bug #18062]
Though this call to `rb_check_type` is just to raise an exception
and never return ac... - 05:51 AM Revision 1a169402 (git): * 2021-08-13 [ci skip]
-
- So that it will not interfere the fast path in `Check_Type`.
-
- Though this call to `rb_check_type` is just to raise an exception
and never return actually, it can return at least formally. That
means a caller function looks like it will access `flags` even in
the special-const cases, and some optim...
08/12/2021
-
- I noticed that `addr2line.c` misinterpreted the size of `DW_FORM_ref_addr` as the size of an address. In fact it should be interpreted as the size of an offset (32 bits in 32-bit DWARF and 64 bits in 64-bit DWARF) [1]. A patch is attache...
-
- Changing state back to open, because this is unresolved and I am repeatedly hitting this issue in various reincarnations. So this is the debug session on PPC64LE:
~~~
$ gdb --args ./miniruby -e'Process.kill("SIGSEGV",$$)'
GNU gdb (G... -
- While it is possible to compile with only the headers, that only works if you know what the headers should contain. If you aren't sure what the headers contain, and you have to test, you really need to test both compiling and linking to...
-
06:16 PM Bug #18073: test/ruby/test_jit.rb: "error: invalid use of '__builtin_va_arg_pack ()'" on Ruby 2.7.4 on gcc 4.8.5
- jaruga (Jun Aruga) wrote in #note-4:
> I executed `git bisect` between `v2_7_3` and `v2_7_4`. And I found the following commit causes these test failures.
> ...
As a temporary workaround to build Ruby 2.7.4 on RHEL7, I am applying a pa... -
- @xtkoba thanks for the info. I confirmed the value of `CPP` on RHEL 7 autoconf version 2.69.
The following `configure.ac` works fine.
```
$ cat configure.ac
AC_INIT
ac_cv_prog_cc_c99=-std=gnu99
AC_PROG_CC_C99
AC_PREPROC_IFELS... -
- The problem described in #note-24 does not look like very much related to the other part. Maybe worth reopening #17052?
-
- Just testing the issue from [comment 24](https://bugs.ruby-lang.org/issues/18062#note-24) and there is not difference with or without `-fno-strict-aliasing`
-
- I was able to reduce the `Hash.ruby2_keywords_hash?(1)` repro and report it to GCC.
It looks like we are hitting an optimizer bug with LTO: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101868
For workarounds, putting `__attribute__ (... -
- vo.x (Vit Ondruch) wrote in #note-26:
> BTW, In your experience, do you think trying something like [1] could help catch such issues earlier?
I'm sorry but I'm not familiar with techniques for detecting strict aliasing violation. It ... -
- xtkoba (Tee KOBAYASHI) wrote in #note-25:
> I would humbly advise that the `-fno-strict-aliasing` flag be appended to `CFLAGS` by default. See also #17540.
I have not find any guidelines about/against `-fno-strict-aliasing` in Fedora... -
01:31 PM Bug #17806: Bad interaction between method cache, prepend, and refinements
- Thanks @alanwu!
I forgot to watch this issue so I was never notified of your response. Sorry for the delay.
> I would try running these two lines before requiring the test file that contains the | call:
> ...
Thank you so much -- ... -
- https://github.com/ruby/reline/commit/040d29b4c0
-
- https://github.com/ruby/irb/commit/98610236ec
-
- https://github.com/ruby/irb/commit/5a0c8dc9e6
-
- This reverts commit c66b27efbb27c6b6b410457282ac841ead1dd132.
-
-
-
02:29 AM Revision 1cdc7510 (git): RBIMPL_ATTR_DEPRECATED: enable for GCC 10.3+
- They fixed the bug.