U++ forum: Welcome to the forum

Search on this site

Search in forums

Home » Community » U++ community news and announcements » 2025.1rc4

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: 2025.1rc4 [message #61600 is a reply to message #61599]

Fri, 21 March 2025 06:09

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

Undefined Behavior Sanitizer.

Attachment: ide.ubsan.01.txt
(Size: 4.84KB, Downloaded 39 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61601 is a reply to message #61600]

Fri, 21 March 2025 06:15

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

On the positive side: a Safe Stack Sanitizer didn't report any problem Smile

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61602 is a reply to message #61601]

Fri, 21 March 2025 07:53

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

Can you send me .bms? (I could figure it out but it would be faster if you did)

Mirek

Report message to a moderator

Re: 2025.1rc4 [message #61603 is a reply to message #61602]

Fri, 21 March 2025 18:20

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Fri, 21 March 2025 02:53

Can you send me .bms? (I could figure it out but it would be faster if you did)

Mirek

I made these bm-files several years ago. Probably, they can be improved.

Attachment: BM.tgz
(Size: 0.62KB, Downloaded 43 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61604 is a reply to message #61603]

Sat, 22 March 2025 02:44

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

I installed missing llvm tools and got better asan and msan logs.
I also attached an improved CLANGcpp17asan.bm.

Attachment: ide.tgz
(Size: 2.86KB, Downloaded 42 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61605 is a reply to message #61604]

Sat, 22 March 2025 08:31

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

Painter/Rasterizer.cpp:318:10: runtime error: left shift of negative value -160

- this is interestingly no more error in C++20 (and ubsan agrees)... (I am adding this note here as I expect you to retest after I am finished fixing).

Report message to a moderator

Re: 2025.1rc4 [message #61606 is a reply to message #61605]

Sat, 22 March 2025 11:30

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

I have fixed everything I could. Unfortunately there are tons of leaks in gtk/fontconfig, unitialized variables etc. outside my control. Also thread sanitizer thinks that malloc/free implementation is wrong Smile

Mirek

Report message to a moderator

Re: 2025.1rc4 [message #61607 is a reply to message #61605]

Sun, 23 March 2025 05:38

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Sat, 22 March 2025 03:31

Painter/Rasterizer.cpp:318:10: runtime error: left shift of negative value -160

- this is interestingly no more error in C++20 (and ubsan agrees)... (I am adding this note here as I expect you to retest after I am finished fixing).

I attached what I'm getting with my compiler.
Left shift is still there ...

Attachment: ide.ubsan.02.txt
(Size: 1.02KB, Downloaded 35 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61608 is a reply to message #61606]

Sun, 23 March 2025 05:43

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Sat, 22 March 2025 06:30

I have fixed everything I could. Unfortunately there are tons of leaks in gtk/fontconfig, unitialized variables etc. outside my control. Also thread sanitizer thinks that malloc/free implementation is wrong

Mirek

Thank you!
IMHO, it makes sense to add build methods with sanitizers to release distros so people won't need to reinvent them.

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61609 is a reply to message #61607]

Sun, 23 March 2025 07:56

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

Novo wrote on Sun, 23 March 2025 05:38

mirek wrote on Sat, 22 March 2025 03:31
Painter/Rasterizer.cpp:318:10: runtime error: left shift of negative value -160

- this is interestingly no more error in C++20 (and ubsan agrees)... (I am adding this note here as I expect you to retest after I am finished fixing).

I attached what I'm getting with my compiler.
Left shift is still there ...

Either your clang version does not know about it or you have not changed the build method to C++20... (When I did, it got fixed).

Mirek

Report message to a moderator

Re: 2025.1rc4 [message #61611 is a reply to message #61609]

Sun, 23 March 2025 20:19

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Sun, 23 March 2025 02:56

Either your clang version does not know about it or you have not changed the build method to C++20... (When I did, it got fixed).

Mirek

Yes, changing cpp version to 20 eliminates the problem.
But current cpp version used by UPP is 17 ...

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61612 is a reply to message #61611]

Sun, 23 March 2025 20:37

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

Novo wrote on Sun, 23 March 2025 20:19

mirek wrote on Sun, 23 March 2025 02:56

Either your clang version does not know about it or you have not changed the build method to C++20... (When I did, it got fixed).

Mirek

Yes, changing cpp version to 20 eliminates the problem.
But current cpp version used by UPP is 17 ...

Well, I suggest you google the issue up... It was defined in practice and widely used for ages now, except the standard lagged behind...

Mirek

Report message to a moderator

Re: 2025.1rc4 [message #61613 is a reply to message #61612]

Mon, 24 March 2025 04:51

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Sun, 23 March 2025 15:37

Well, I suggest you google the issue up... It was defined in practice and widely used for ages now, except the standard lagged behind...

Mirek

I propose a different way of fixing this problem: a pull request.
New command to check: umk tutorial CoreTutorial CLANGcpp17ubsan -bus
Result is attached.

Attachment: CoreTutorial.ubsan.01.txt
(Size: 0.62KB, Downloaded 43 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61614 is a reply to message #61613]

Mon, 24 March 2025 08:08

mirek
Messages: 14263
Registered: November 2005

Ultimate Member

Novo wrote on Mon, 24 March 2025 04:51

mirek wrote on Sun, 23 March 2025 15:37

Well, I suggest you google the issue up... It was defined in practice and widely used for ages now, except the standard lagged behind...

Mirek

I propose a different way of fixing this problem: a pull request.
New command to check: umk tutorial CoreTutorial CLANGcpp17ubsan -bus
Result is attached.

a) good to know

b) you are adding a couple of completely unnecessarry and confusing lines to the project

c) AFAIK you can as well do -fsanitize=no-shift-base if you insist on using C++17 definition (but why you insist on that is beyond me. we will probably be moving C++20 in 3 years anyway)

Report message to a moderator

Pages (3): [ « ‹ 1 2 3]

Previous Topic:	2025.1rc3
Next Topic:	2025.1rc5

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Tue Jul 01 19:42:49 CEST 2025

Total time taken to generate the page: 0.03004 seconds