Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <58068fc813f1c55d2b3dde305d2b8e29e9801668.camel@zv.io>
Date: Tue, 19 Jan 2021 17:01:28 -0600
From: Zach van Rijn <me@...io>
To: Rasmus Andersson <rasmus@...ion.se>
Cc: musl@...ts.openwall.com
Subject: Re: waitpid (wait4) on Linux 5 returns invalid values

On Tue, 2021-01-19 at 14:02 -0800, Rasmus Andersson wrote:
> The program compiles correctly with
> https://more.musl.cc/9/x86_64-linux-musl/x86_64-linux-musl-native.tgz
> (The GCC 9 version) Perhaps an issue with GCC 10?

Rasmus, thank you for reporting this issue.


Using your reproducer I'm able to observe the following:

(1) GCC 10 as-published on musl.cc : BUG
waitpid returned -10, wstat=0, errno=0 (No error information)

(2) GCC 10 without this patch [1]  : OK
waitpid returned -1, wstat=0, errno=10 (No child process)

which matches your GCC 9 (20200828) observation.


This patch was applied to musl.cc as part of a series to add
experimental riscv32 support in September 2020, but was applied
universally and not strictly riscv32 targets.

The patch in question was /not/ applied to the GCC 9 binaries as
they were last updated in August of 2020, nor to earlier 10 ones.

I am updating the build infrastructure to avoid contamination of
"supported" targets by experimental patches such as this one.

New (fixed and newer GCC) toolchains will be published as soon as
they finish building (~24 hours). I will ping you off-list.

Additional tests (not simply spot-checks) will be implemented for
future releases.


Sorry for the inconvenience.


[1]: https://www.openwall.com/lists/musl/2020/09/03/14

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.