[Bug 46273] New: rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

Bug ID: 46273
           Summary: rand() doesn't get a unique state per-fiber like it
                    appears to on native windows
           Product: Wine
           Version: 1.8.7
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: winecrt0
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

I've been using WINE to test windows builds during game development in Debian
GNU/Linux.

I use the windows fibers API to implement coroutines and in testing on Win7
noticed a bunch of coroutine-backed game objects using rand() for randomized
visible things were unexpectedly looking uniform.  In WINE they had all
appeared distinctly random.

I'm not an experienced windows developer, but it /appears/ that in Win7,
calling rand() in a newly created fiber restarts from the default seed rather
than sharing the same PRNG state with the other fibers.  It's rather odd, but I
verified it with a smaller test program that only printed the return value from
rand() in new fibers, and they were all printing the same value.

It's possible this is already fixed in a newer WINE version, I'm limited to
what's bundled in debian 9.5 (1.8.7).  Or maybe this is considered a Windows
bug and WINE's behavior is deliberately different.

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

--- Comment #1 from Austin English  ---
As I'm sure you realize, 1.8.x is very outdated (we're already in the 4.0
release process). If you could test from a newer debian repo (i.e.,
testing/unstable) that would be helpful.

In any case, could you attach the test source/binary?

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

--- Comment #2 from [e-mail] ---
Created attachment 62984
  --> https://bugs.winehq.org/attachment.cgi?id=62984
rand() from fibers test program

This program creates 100 fibers and waits for their exit.

The fibers all execute the same function, in pseudo-code:
{
        print(rand())
        sleep_schedule_others(1s)
        print(rand())
}

In WINE 1.8.7, I observe all the fibers printing random numbers.

In Win7, I observe 100 of the same random number, a delay, then 100 of another
uniform random number.

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

joaopa  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #3 from joaopa  ---
As already said, what is the purpose to test wit such an old version of wine?
Test with wine-4.0-rc1.

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1

--- Comment #4 from Nikolay Sivov  ---
(In reply to joaopa from comment #3)
> As already said, what is the purpose to test wit such an old version of wine?
> Test with wine-4.0-rc1.

And now you can try it yourself with attached test program.

(In reply to winebz from comment #2)
> Created attachment 62984 [details]
> rand() from fibers test program
>
> This program creates 100 fibers and waits for their exit.
>
> The fibers all execute the same function, in pseudo-code:
> {
>         print(rand())
>         sleep_schedule_others(1s)
>         print(rand())
> }
>
> In WINE 1.8.7, I observe all the fibers printing random numbers.
>
> In Win7, I observe 100 of the same random number, a delay, then 100 of
> another uniform random number.

Yes, I can confirm.

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

--- Comment #5 from [e-mail] ---
If it's of any use, I've made the source available via giet in a
semi-convenient to build form (autotools):

`git clone git://git.pengaru.com/randtest`

See the README for instructions on building with mingw.  Please note there are
submodules requiring a recursive init for the fibers support, as explained in
the README.

I only wanted to document the apparent difference from genuine Windows in
filing this bug.  For my purposes I've switched to different PRNG, it's not
important to fix from my perspective.

It might make the most sense to simply document the bug.

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download, source, testcase

[Bug 46273] rand() doesn' t get a unique state per-fiber like it appears to on native windows

https://bugs.winehq.org/show_bug.cgi?id=46273

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
       Distribution|---                         |Debian