InputTextFilterCharacter doesn't filter Tabs (in german system)

[HaSa1002]

Version/Branch of Dear ImGui:

Version: 1.69
Branch: master

Back-end/Renderer/Compiler/OS

Back-ends: imgui_sfml
Operating System: Windows 10 (German as Systemlanguage)

My Issue:
You can enter a Tab in any InputText, which will cause a focus jump to the next item, but the tab is also written into the buffer. My tests of the InputTextFilterCharacter (which is my best friend...) proved, that isprint returns 64 if a tab is entered. Why is it necessary to use isprint in that case?

Standalone, minimal, complete and verifiable example:

ImGui::Begin("Bug");
static char buff[5] = "";
static char buff2[5] = "";
ImGui::InputText("Label", buff, 5);
ImGui::InputText("Label2", buff2, 5);
ImGui::End();

[ocornut]

I am not sure the SFML back-end is correct emitting both a Key and Char input.

[HaSa1002]

How can I check that?

[ocornut]

How can I check that?

Add print out into the SFML code calling io.AddInputCharacter() and setting KeysDown[] = xxx value. In the printout include the value of ImGui::GetFrameCount() so you can see the frame timing at which each is being emitted.

But I was wrong: most back-ends are emitting a \t character along with the Tab key. Also see #1336, so this isn't the source of the problem.

that isprint returns 64 if a tab is entered.

Could you compile this code and confirm the output:

#include <ctype.h>
[...]
    printf("isprint \\t = %d\n", isprint('\t'));
    printf("isprint \\n = %d\n", isprint('\n'));

Then provide the exact compiler you are using (which you have omitted from the issue template questions).

Thank you.

[ocornut]

Alright, welcome to compiler/std-library hell :)

https://stackoverflow.com/questions/51435249/isprint-t-evaluates-to-true-64-with-md-compiler-option
and
JuliaLang/julia#7416

I will explicitely work-around around this in the code, thanks!

[ocornut]

Fixed this now, thank you for the report! Also probably explains #1336 now!

[codename-irvin]

FYI, I am still seeing this issue on Win32/DX12. I have worked around the issue for now by commenting out this code.

[ocornut]

I can confirm there's an issue with the new IO queue trickling events over two frames (#4921) those tests are now invalid.

[ocornut]

@codename-irvin can you clarify precisely the issue you are getting?
I am getting two tabs inserted instead of one.
Are you on a 1.87+ backend?

[ocornut]

Fixed by dfbe938

[codename-irvin]

@ocornut, indeed. The issue was that I was getting two tabs inserted instead of one. Looks like I am on a 1.88 backend. I am on docking branch latest. Thanks for the fix!