[Bug target/92462] [arm32] -ftree-pre makes a variable to be wrongly hoisted out

[aleksei.voitylov@bell-sw.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92462

--- Comment #9 from Aleksei Voitylov <aleksei.voitylov@bell-sw.com> ---
(In reply to Alexander Monakov from comment #8)
> The full preprocessed source is provided and it clearly says
> 
>     typedef unsigned char uint8_t;
> 
> in line 10, so it is in fact a character type.
> 
> It's suspicious that cmpxchg_using_helper does not return a value
> (incorrectly reduced testcase?) and there's still an aliasing violation when
> atomic_cmpxchg_func tries to cast 'dest' from uint8_t* to int*. I think the
> report was closed prematurely.
It is suspicious regarding atomic functionality - from that point the testcase
is over-reduced, you are right. It doesn't invalidate the initial question
about the  hoisting, though.

> Aleksei - always provide output of 'gcc -v' when reporting such bugs,
> otherwise people may be unable to reproduce it when there's really a problem
> (no way to tell how your compiler was configured or even its exact version).

$ g++ -v
Using built-in specs.
COLLECT_GCC=/home/build/toolchain/gcc-linaro-7.3.1-2018.05-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-g++
COLLECT_LTO_WRAPPER=/home/build/toolchain/gcc-linaro-7.3.1-2018.05-x86_64_arm-linux-gnueabihf/bin/../libexec/gcc/arm-linux-gnueabihf/7.3.1/lto-wrapper
Target: arm-linux-gnueabihf
Configured with:

<skipped>

Thread model: posix
gcc version 7.3.1 20180425 [linaro-7.3-2018.05 revision
d29120a424ecfbc167ef90065c0eeb7f91977701] (Linaro GCC 7.3-2018.05)