[Bug 49496] New: sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

Bug ID: 49496
           Summary: sprintf uses locale specific decimal point when it
                    should not.
           Product: Wine
           Version: 5.11
          Hardware: x86
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

This manifested first in Microsoft Word 2007 printing
empty pages with wine-5.11.
I could track this down to wineeps.drv using
my locales "de_DE.UTF-8" decimalpoint ','
instead of the expected '.' from the "C" locale:

- /FontBBox [-0.476074 -0.193848 1.213867 0.952148] def
   + /FontBBox [-0,476074 -0,193848 1,213867 0,952148] def

This line is produced in wineps.drv\type42.c:207, which is surrounded
by setlocale calls to use "C". But these do call into MSVCRT_setlocale,
which seems not forward this into glibc. And in the end the "%f" format
is forwarded to glibc's __sprintf.

A git bisect led to this commit:

b86dc3926bfe5cd92400aa96c89b0255eba1d447 is the first bad commit
    commit b86dc3926bfe5cd92400aa96c89b0255eba1d447
    Author: Alexandre Julliard 
    Date:   Thu Jun 11 11:38:16 2020 +0200

ntdll: Move the locales initialization to the Unix library.

Signed-off-by: Alexandre Julliard

This, I guess, is also visible in the tests of ucrtbase:printf since
that day, where some machines seem to use french locale:

https://test.winehq.org/data/tests/ucrtbase:printf.html

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

Piotr Caban  changed:

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

--- Comment #1 from Piotr Caban  ---
FYI I'm working on changing the msvcrt printf implementation so it doesn't use
sprintf from glibc for outputing floating point numbers. It still needs quite a
lot of work.

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression
    Regression SHA1|                            |b86dc3926bfe5cd92400aa96c89
                   |                            |b0255eba1d447

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

--- Comment #2 from Bernhard Übelacker  ---
Did not test it yet, but I guess commit 031c744a1fb9 might fix this issue,
until sprintf from glibc is no longer used.
And then this bug might be also a duplicate of #49489.

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

Bernhard Übelacker  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|UNCONFIRMED                 |RESOLVED
      Fixed by SHA1|                            |031c744a1fb9b94f7a4634eae5f
                   |                            |56845b586554d

--- Comment #3 from Bernhard Übelacker  ---
Found this could be reproduced also with just printing in notepad.

Additionally can confirm that 031c744a1fb9 indeed fixed the issue,
therefore marking as resolved.

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

--- Comment #4 from Rafał Mużyło  ---
...ah, shrub, it isn't resolved yet.

If you'd read julliard's post on the list, you'd know that the workaround was
green-lighted only because the proper solution was taking too much time.

Only once that is written and goes in we'll be in position to know how many
regressions it will trigger.

Also, besides printf, its quite likely also scanf will need to be addressed.

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

--- Comment #5 from Bernhard Übelacker  ---
I opened this bug report as regression caused by commit b86dc3926bfe5
which got mitigated by commit 031c744a1fb9.
Therefore I marked this report as fixed and I guess it will be
up to Piotr to decide if he needs/wants this report stay open?

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

--- Comment #6 from Piotr Caban  ---
(In reply to Rafał Mużyło from comment #4)
> Also, besides printf, its quite likely also scanf will need to be addressed.

Scanf is already not depending on glibc locale.

[Bug 49496] sprintf uses locale specific decimal point when it should not.

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49496

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #7 from Alexandre Julliard  ---
Closing bugs fixed in 5.13.