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 #61572 is a reply to message #61568]

Tue, 18 March 2025 23:00

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Tue, 18 March 2025 06:21

I do not observe any of these ill effects...

It is reproducible in my case. Sometimes I need to call TheIDE several times till it friezes.

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61573 is a reply to message #61569]

Wed, 19 March 2025 00:09

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

Klugier wrote on Tue, 18 March 2025 14:26

Hello,

I would like to jump to the discussion. Do we want to support macOS 10.15, if it is currently not supported by Apple? The support for this OS ended on 12 Sep 2022, which is 2 and the half years ago. For me, we should focus only on macOS that is currently supported by the Apple. We don't have enough manpower for that. Also, Intel CPU's are supported on the latest version of macOS, which is macOS Sequoia (15.0). So, we will do not drop architecture, but only unsupported versions.

BTW, There is a compilation issue on our pipelines. We should focus to fix this issue...

Sources:
- https://endoflife.date/macos

Klugier

Well, a lot of people (including me) use old hardware. I even bought a server rack to put all my old servers into it Smile

Used professional hardware is very good and cheap Smile

I bought a brand new iMac for my wife. It broke in five years. I bought another brand new iMac. It broke in five years as well. After that I decided not to waste my money on Apple's hardware and bought an old iMac off ebay for a tiny fraction of price of a new one. I also bought a spare one because I didn't expect the first one to last for very long. Interestingly, the first one still works and it runs macOS 10.13 Smile

And I have an old MacBook Pro which runs macOS 10.15. I do not need anything better than this MacBook. It has 16GB of RAM and runs photo processing software I use Smile

Well, I also have a spare iMac which I bought for my wife Smile

So, support of the old hardware is important.
I also thinking about using of U++ for IoT projects. These days C++ became new assembler Smile

Regards,
Novo

[Updated on: Wed, 19 March 2025 00:12]

Report message to a moderator

Re: 2025.1rc4 [message #61574 is a reply to message #61569]

Wed, 19 March 2025 10:02

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Klugier wrote on Tue, 18 March 2025 19:26

Hello,

I would like to jump to the discussion. Do we want to support macOS 10.15, if it is currently not supported by Apple? The support for this OS ended on 12 Sep 2022, which is 2 and the half years ago. For me, we should focus only on macOS that is currently supported by the Apple. We don't have enough manpower for that. Also, Intel CPU's are supported on the latest version of macOS, which is macOS Sequoia (15.0). So, we will do not drop architecture, but only unsupported versions.

BTW, There is a compilation issue on our pipelines. We should focus to fix this issue...

Sources:
- https://endoflife.date/macos

Klugier

It is true, however I think it is wise to pursue this, at least for the "hidden menu issue", as there is no obvious reason for it not to work and so it can be a hidden bug (that would strike back later in some corner case).

Report message to a moderator

Re: 2025.1rc4 [message #61575 is a reply to message #61570]

Wed, 19 March 2025 10:39

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Tue, 18 March 2025 22:24

mirek wrote on Tue, 18 March 2025 06:21

~/.local/state/u++/log/...

This is how it looks in GUI ...
It turned out that a log-file is created, but it is located in [root of .config]/.local/state/u++/log/...
And TheIDE looks for a log file in a fixed place and ignores GetFileFolder(GetUserConfigDir()) ...

Well, it appears you might have active "sandboxing feature" (that is, you have .config in some dir that is parent of executable dir and is not ~/.config). Which is something that was not accounted for in TheIDE. Tried to fix it...

https://github.com/ultimatepp/ultimatepp/commit/4c6c1d701c77 d3622c51767e39c2f7cf754a1cf6

Report message to a moderator

Re: 2025.1rc4 [message #61576 is a reply to message #61564]

Wed, 19 March 2025 10:59

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Mon, 17 March 2025 22:53

mirek wrote on Mon, 17 March 2025 15:18
Wait a moment: Does it mean that what is missing is texts? I mean, that submenus are under "spaces"?

Mirek

That is correct. As I wrote previously: "It works, but it is invisible.". Probably, I should've described the problem in more details.

OK, .log ruled out some issues, so I have tried to fix something else:

https://github.com/ultimatepp/ultimatepp/commit/2c64e6cb50a0 5dba0eae893c3985ee139a7fd25d

Please check.

Report message to a moderator

Re: 2025.1rc4 [message #61577 is a reply to message #61576]

Wed, 19 March 2025 18:39

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Wed, 19 March 2025 05:59

Please check.

Menu works. Viewing of log-files with TheIDE works. TheIDE doesn't seem to hang anymore.

Thank you!

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61579 is a reply to message #61577]

Wed, 19 March 2025 20:38

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Wed, 19 March 2025 18:39

mirek wrote on Wed, 19 March 2025 05:59
Please check.

Menu works. Viewing of log-files with TheIDE works. TheIDE doesn't seem to hang anymore.

Thank you!

Thats a good news! Thanks for testing and patience.

Report message to a moderator

Re: 2025.1rc4 [message #61580 is a reply to message #61579]

Thu, 20 March 2025 05:47

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

I checked TheIDE with valgrind (debug version of TheIDE + open the ide package). A result is attached.
I ran valgrind three times and each time result was different.
This doesn't look good to me. I doubt that this is a problem with clangd itself.

Attachment: valgrind2.log
(Size: 12.77KB, Downloaded 42 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61581 is a reply to message #61580]

Thu, 20 March 2025 06:17

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

Interestingly, I tried to compile TheIDE with a USEMALLOC flag and got a result below.

Linking...
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Util.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
...

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61584 is a reply to message #61581]

Thu, 20 March 2025 08:58

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Thu, 20 March 2025 06:17

Interestingly, I tried to compile TheIDE with a USEMALLOC flag and got a result below.

Linking...
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Util.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
...

There was a related bug in util.cpp, but not this one...

Anyway, after fixing it, I have checked on all 3 platforms. So I guess please retry with master.

Report message to a moderator

Re: 2025.1rc4 [message #61585 is a reply to message #61580]

Thu, 20 March 2025 09:17

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Thu, 20 March 2025 05:47

I checked TheIDE with valgrind (debug version of TheIDE + open the ide package). A result is attached.
I ran valgrind three times and each time result was different.
This doesn't look good to me. I doubt that this is a problem with clangd itself.

Checked our code, I do not see anything wrong with it...

And it is really simple really:

	CXCursor cursor = clang_getTranslationUnitCursor(tu);
	pp_id = clang_getCursorPrintingPolicy(cursor);
	pp_pretty = clang_getCursorPrintingPolicy(cursor);
	for(int i = 0; i <= CXPrintingPolicy_LastProperty; i++) {
		clang_PrintingPolicy_setProperty(pp_id, (CXPrintingPolicyProperty)i, 0);
		clang_PrintingPolicy_setProperty(pp_pretty, (CXPrintingPolicyProperty)i, 0);
	}

	for(CXPrintingPolicyProperty p : {
			CXPrintingPolicy_SuppressSpecifiers,
			CXPrintingPolicy_SuppressTagKeyword,
			CXPrintingPolicy_SuppressUnwrittenScope,
			CXPrintingPolicy_SuppressInitializers,
			CXPrintingPolicy_SuppressStrongLifetime,
			CXPrintingPolicy_SuppressLifetimeQualifiers,
			CXPrintingPolicy_SuppressTemplateArgsInCXXConstructors,
			CXPrintingPolicy_TerseOutput,
			CXPrintingPolicy_SuppressImplicitBase,
#ifndef UBUNTU2204_WORKAROUND
			CXPrintingPolicy_FullyQualifiedName,
#endif
			CXPrintingPolicy_Bool })
		clang_PrintingPolicy_setProperty(pp_id, p, 1);

	clang_PrintingPolicy_setProperty(pp_pretty, CXPrintingPolicy_TerseOutput, 1);
	clang_PrintingPolicy_setProperty(pp_pretty, CXPrintingPolicy_Bool, 1);
	clang_PrintingPolicy_setProperty(pp_pretty, CXPrintingPolicy_SuppressScope, 1);
	initialized = true;
	clang_visitChildren(cursor, clang_visitor, this); <<<<<<<<<<<<<<<< valgrind problem

So the only input is 'tu', the rest are legal libclang calls. And checking the rest of code, 'tu' is either nullptr or result of translate....

Mirek

Report message to a moderator

Re: 2025.1rc4 [message #61588 is a reply to message #61585]

Thu, 20 March 2025 17:20

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Thu, 20 March 2025 04:17

Checked our code, I do not see anything wrong with it...

In such case this is a problem with clangd Smile

I tried to double-check that with CLANG sanitizers but bumped into a problem with USEMALLOC ...

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61589 is a reply to message #61584]

Thu, 20 March 2025 17:23

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Thu, 20 March 2025 03:58

Anyway, after fixing it, I have checked on all 3 platforms. So I guess please retry with master.

Below is what I'm getting now.

Linking...
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Util.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `Upp::SslFree(void*, char const*, int)':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/InitExit.cpp:41:(.text+0x3d7): undefined reference to `Upp::GetMemoryBlockSize(void*)'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/InitExit.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Socket.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Socket.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Socket.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/P7S.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/P7S.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/P7S.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/SSL.icpp.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/SSL.icpp.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/SSL.icpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
clang++: error: linker command failed with exit code 1 (use -v to see invocation)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61593 is a reply to message #61589]

Thu, 20 March 2025 23:10

mirek
Messages: 14261
Registered: November 2005

Ultimate Member

Novo wrote on Thu, 20 March 2025 17:23

mirek wrote on Thu, 20 March 2025 03:58

Anyway, after fixing it, I have checked on all 3 platforms. So I guess please retry with master.

Below is what I'm getting now.

Linking...
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Util.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Util.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `Upp::SslFree(void*, char const*, int)':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/InitExit.cpp:41:(.text+0x3d7): undefined reference to `Upp::GetMemoryBlockSize(void*)'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/InitExit.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/InitExit.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Socket.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/Socket.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/Socket.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/P7S.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/P7S.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/P7S.cpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/SSL.icpp.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/Core.h:426:(.text.startup+0xc): undefined reference to `MemDiagCls::MemDiagCls()'
/bin/x86_64-unknown-linux-gnu-ld: /home/xxx/dvlp/cpp/code/upp/out/Core/SSL/CLANGcpp17um.Debug.Debug_Full.Gui.Shared/SSL.icpp.o: in function `__cxx_global_var_init':
/home/xxx/dvlp/cpp/code/upp/git/uppsrc/Core/SSL/SSL.icpp:(.text.startup+0x13): undefined reference to `MemDiagCls::~MemDiagCls()'
clang++: error: linker command failed with exit code 1 (use -v to see invocation)

There must be some difference. What exactly is the host / build method / settings?

Report message to a moderator

Re: 2025.1rc4 [message #61594 is a reply to message #61593]

Fri, 21 March 2025 02:39

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

mirek wrote on Thu, 20 March 2025 18:10

There must be some difference. What exactly is the host / build method / settings?

Host: Void Linux glibc (this is a rolling distro). Clang v17.0.6.
Build method: attached
Build command: umk uppsrc ide CLANGcpp17um -busa
umk version: 17795

This build method was made long time ago, but it looks OK to me.

Attachment: CLANGcpp17um.bm
(Size: 0.64KB, Downloaded 36 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61595 is a reply to message #61594]

Fri, 21 March 2025 04:52

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

It turned out that address sanitizer doesn't require USEMALLOC ...
Compilation command: umk uppsrc ide CLANGcpp17asan -busa
A result is attached.

EDITED: I guess this error happens because USEMALLOC is still required ...

Attachment: ide.asan.01.txt
(Size: 4.43KB, Downloaded 26 times)

Regards,
Novo

[Updated on: Fri, 21 March 2025 05:05]

Report message to a moderator

Re: 2025.1rc4 [message #61596 is a reply to message #61595]

Fri, 21 March 2025 05:11

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

Well, replacing 'COMMON_FLAGS = ".USEMALLOC";' with 'COMMON_FLAGS = "USEMALLOC";' fixed problem with linking.
Weird.

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61597 is a reply to message #61596]

Fri, 21 March 2025 05:21

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

This is what I've got after I enabled USEMALLOC with address sanitizer.

Attachment: ide.asan.02.txt
(Size: 12.44KB, Downloaded 29 times)

Regards,
Novo

Report message to a moderator

Re: 2025.1rc4 [message #61598 is a reply to message #61597]

Fri, 21 March 2025 05:43

Novo
Messages: 1430
Registered: December 2006

Ultimate Contributor

Memory sanitizer. This one looks weird because sanitizers are able to instrument only code they compile (unlike valgrind).
But still. It is what it is.