Re: 10.99.12 and 9.3 both issue (Re: ruby 3.4.x Segfaults issue

[Ryo ONODERA <ryo%tetera.org@localhost>]

Hi,

Takahiro Kambe <taca%NetBSD.org@localhost> writes:

> In message <87wmdavhig.wl-makoto%if.t.u-tokyo.ac.jp@localhost>
> 	on Fri, 28 Feb 2025 13:07:51 +0900,
> 	Makoto Fujiwara <makoto%if.t.u-tokyo.ac.jp@localhost> wrote:
>> I've tried the similar thing on NetBSD/amd64 9.3,
>> and I've  got the (probably) exact results.
> I found out what is happennig.  This problem exists not only NetBSD
> current but 10.1_STABLE (and 9.x).
>
> The problem exists in PRISM parser, Ruby 3.4 use it as default parser.
> Old parser is still supported, using "--parser=parse.y" option solves
> this problem.
>
> (PRISM parser was introduced on Ruby 3.3 but it was experimental and
> was not default parser.  It could not be used for reading from stadard
> input.)
>
> Line 22617 from prism/prism.c:
>
> static bool
> pm_parse_stream_read(pm_buffer_t *buffer, void *stream, pm_parse_stream_fgets_t *fgets) {
> #define LINE_SIZE 4096
>     char line[LINE_SIZE];
>
>     while (memset(line, '\n', LINE_SIZE), fgets(line, LINE_SIZE, stream) != NULL) {
>
>
> This code expects call third parameter of pm_parse_stream_read() but
> it calls standard fgets(3) as mef@ showed.
>
> In message <87y0xqvib0.wl-makoto%if.t.u-tokyo.ac.jp@localhost>
> 	on Fri, 28 Feb 2025 12:50:43 +0900,
> 	Makoto Fujiwara <makoto%if.t.u-tokyo.ac.jp@localhost> wrote:
>> |  (gdb) bt
>> |  #0  0x00007dc205dd0961 in memchr () from /usr/lib/libc.so.12
>> |  #1  0x00007dc205d942ed in fgets () from /usr/lib/libc.so.12
>> |  #2  0x00007dc2063bb0d9 in pm_parse_stream_read.constprop () from /usr/pkg/lib/libruby34.so.34
>> |  #3  0x00007dc2063ec6df in pm_parse_stream () from /usr/pkg/lib/libruby34.so.34
>> |  #4  0x00007dc20619b32e in pm_parse_stdin () from /usr/pkg/lib/libruby34.so.34
>> |  #5  0x00007dc2062cfa7f in process_options () from /usr/pkg/lib/libruby34.so.34
>
> This dose not happen on macOS 15.3.1 and maybe other systems and
> I don't know what is the proper way to solve by compiler framework.
>
> Anyway, I don't think it is *good* habbit to write such codes.
>
> ruby34-3.4.2nb1 will be solve the problem.

Under my NetBSD/amd64-current of yesterday, lang/ruby34-3.4.2nb1
works fine now.
Thank you very much!!!

> Best regards.
>
> -- 
> Takahiro Kambe <taca%NetBSD.org@localhost>

-- 
Ryo ONODERA // ryo%tetera.org@localhost
PGP fingerprint = 82A2 DC91 76E0 A10A 8ABB  FD1B F404 27FA C7D1 15F3