[WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace #3003

ErnyTech · 2020-03-24T18:46:36Z

This allows you to view demangled and pretty printed C++ symbols
in the stacktrace.
Is especially useful considering that D is getting more and more support for C++.

Now the C++ symbols in the stacktrace will be printed like this:

        test.d:7 [C++] Test<int>::SomeName(int, long, int) [0x55711efee36f]
        test.d:18 [C++] testcpp() [0x557228e17978]

This need the PR #3002

Geod24 · 2020-03-27T04:44:27Z

FYI we now have dlang/dmd#10964 in DMD.
We might be interested in having it here, too.

rainers · 2020-03-27T07:11:02Z

FYI there was a similar PR: #2083

I think this one has better separation. For Windows, the VC mangling should be supported via UnDecorateSymbolName. Please also add a test case.

rainers · 2020-03-27T07:15:08Z

src/core/sys/windows/stacktrace.d

-        res ~= demangle(tempSymName, demangleBuf);
+        auto demangledName = demangle(tempSymName, demangleBuf);
+
+        version (Shared)


Why limit this to a shared build of the runtime?

you're right, for Windows it is not necessary

Why is it necessary on other platforms? "Shared" is not a version for dlopen support, but whether the runtime itself is a shared library.

druntime must be linked with -ldl, otherwise the linker will not find dlopen.
I don't know how i can check this in a better way

The compiler already adds -ldl to the linker command line (at least for linux):
https://github.com/dlang/dmd/blob/e8f984995598543279c07b9800d19a9e19321562/src/dmd/link.d#L756

Ok but if I create the object files with dmd and then link it manually? I'm not sure it's a good idea to depend on libdl for druntime

Ah, I see that rt_loadLibrary is only implemented if version(Shared), so it's consistent to not use it here, too. That restriction seems a bit strange, though.

Maybe we could do better using weak symbols.
It would be enough to create a dlopen function that returns null marked as weak symbol, so if the user links libdl the real dlopen would be used

Geod24

Added a few comments about the second commit

Geod24 · 2020-04-02T01:54:24Z

changelog/cpptrace.dd

@@ -0,0 +1,51 @@
+Added experimental `C++` symbol demanling in the stacktrace


Suggested change

Added experimental `C++` symbol demanling in the stacktrace

The runtime can now demangle C++ names in stack traces

A bit more selling to the user (we tend to mark everything as "experimental" and that doesn't inspire confidence).
You can just note later that it is not enabled by default.

Geod24 · 2020-04-02T02:08:33Z

changelog/cpptrace.dd

@@ -0,0 +1,51 @@
+Added experimental `C++` symbol demanling in the stacktrace
+
+This feature is availableby  passing the following switches to your executable (not to the compiler):


Ideally, the release note should start by describing the feature itself. You've been working on it for quite some time, so you know exactly what it entails, but not everyone does. Something like:

When a stack trace was generated by the runtime (e.g. when calling `Throwable.toString`), C++ names were not previously demangled. Hence the following code: --- <INSERT CODE> --- Would lead to the following stack trace: --- <INSERT MANGLED STACK TRACE> --- Starting from this release, C++ names will be demangled whenever possible, hence the previous stack trace will now show as: --- <INSERT DEMANGLED STACK TRACE> --- This feature is still experimental and must be opted-in by <BLAH BLAH>. Demangling is only possible when <BLAH BLAH>.

Note that this is just an example, but it illustrate a few points:

Always describe the feature first, what it brings to the user: imagine giving an elevator pitch about it;

Try to show the difference between the old behavior and the new one: remember users might not be familiar with the old behavior, especially if someone looks at this release note in a few year;

Include a code example whenever possible (which you did already);

Put restrictions / limitations at the end;

Geod24 · 2020-04-02T02:10:05Z

src/core/internal/cpptrace.d

+
+private struct CPPTrace
+{
+    __gshared CPPDemangle instance;


Don't we already have a singleton in the other module?

Geod24 · 2020-04-02T02:10:53Z

src/core/internal/cpptrace.d

+    __gshared CPPDemangle instance;
+    __gshared Config config;
+
+    static this()


A static this() is run on each thread constructions. You want a shared static this here, no ?

Having a C++ symbol demangler is very useful, with this it will be possible to implement the C++ symbol demangle in the stacktrace. The demangling is done in Posix platform by function __cxa_demangle presents in Itanium C++ ABI (https://itanium-cxx-abi.github.io/cxx-abi/abi.html#demangler). Itanium C++ ABI is used practically in all modern C++ compilers on Posix, however old compilers (like GCC < 3.0) used a different name mangling, this is not a problem because DMD in fact only supports Itanium C++ ABI on Posix. For Windows instead the implementation it is provided by the UnDecorateSymbolName function present in the Debug Help Library. (https://docs.microsoft.com/en-us/windows/win32/api/dbghelp/nf-dbghelp-undecoratesymbolname) Signed-off-by: Ernesto Castellotti <[email protected]>

This allows you to view demangled and pretty printed C++ symbols in the stacktrace. Is especially useful considering that D is getting more and more support for C++. Now the C++ symbols in the stacktrace will be printed like this: test.d:7 [C++] Test<int>::SomeName(int, long, int) [0x55711efee36f] test.d:18 [C++] testcpp() [0x557228e17978] Signed-off-by: Ernesto Castellotti <[email protected]>

RazvanN7 · 2022-01-21T08:22:20Z

@ErnyTech What's the status on this?

ErnyTech force-pushed the cppstack branch from dfb01dc to 5022f16 Compare March 26, 2020 14:47

ErnyTech changed the title ~~Enable demangling of C++ symbols in the stacktrace~~ [WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace Mar 26, 2020

ErnyTech force-pushed the cppstack branch 8 times, most recently from df6a61a to 3e03206 Compare March 26, 2020 17:00

ErnyTech mentioned this pull request Mar 26, 2020

Add cppdemangle in druntime #3002

Open

rainers reviewed Mar 27, 2020

View reviewed changes

ErnyTech force-pushed the cppstack branch 13 times, most recently from 9ccee09 to 18fd3ec Compare April 1, 2020 12:34

Geod24 reviewed Apr 2, 2020

View reviewed changes

ErnyTech force-pushed the cppstack branch from 18fd3ec to 750ce1d Compare April 12, 2020 19:38

ErnyTech force-pushed the cppstack branch from 750ce1d to d4c40a5 Compare April 12, 2020 19:43

dlang-bot added Needs Rebase needs a `git rebase` performed Needs Work stalled labels May 27, 2021

dlang-bot removed the stalled label Jan 21, 2022

dlang-bot added the stalled label Apr 21, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace #3003

[WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace #3003

ErnyTech commented Mar 24, 2020 •

edited

Loading

Geod24 commented Mar 27, 2020

rainers commented Mar 27, 2020

rainers Mar 27, 2020

ErnyTech Mar 27, 2020

rainers Mar 27, 2020 •

edited

Loading

ErnyTech Mar 27, 2020 •

edited

Loading

rainers Mar 27, 2020

ErnyTech Mar 27, 2020

rainers Mar 27, 2020

ErnyTech Mar 27, 2020 •

edited

Loading

Geod24 left a comment

Geod24 Apr 2, 2020

Geod24 Apr 2, 2020

Geod24 Apr 2, 2020

Geod24 Apr 2, 2020

RazvanN7 commented Jan 21, 2022

		@@ -0,0 +1,51 @@
		Added experimental `C++` symbol demanling in the stacktrace

	Added experimental `C++` symbol demanling in the stacktrace
	The runtime can now demangle C++ names in stack traces

		@@ -0,0 +1,51 @@
		Added experimental `C++` symbol demanling in the stacktrace

		This feature is availableby passing the following switches to your executable (not to the compiler):

[WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace #3003

Are you sure you want to change the base?

[WIP] [DONT MERGE] Enable demangling of C++ symbols in the stacktrace #3003

Conversation

ErnyTech commented Mar 24, 2020 • edited Loading

Geod24 commented Mar 27, 2020

rainers commented Mar 27, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rainers Mar 27, 2020 • edited Loading

Choose a reason for hiding this comment

ErnyTech Mar 27, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ErnyTech Mar 27, 2020 • edited Loading

Choose a reason for hiding this comment

Geod24 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

RazvanN7 commented Jan 21, 2022

ErnyTech commented Mar 24, 2020 •

edited

Loading

rainers Mar 27, 2020 •

edited

Loading

ErnyTech Mar 27, 2020 •

edited

Loading

ErnyTech Mar 27, 2020 •

edited

Loading