This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: New demangler in C++


>When Alex first wrote this code, it was incorporated into the library. 
>Then, the FSF refused to license the demangler under terms that
>permitted that, so we removed it.  They then agreed to license the
>demangler under terms that would permit that, but the code never went
>back into the library.

__cxa_demangle should only be defined as part of libsupc++. 

>The key point is that there are several thousand lines of code in there
>that do exactly the same stuff as the new code in demangle.h.

.. aaah but without the advantages I mentioned in my first post. The two
demanglers are designed very differently, for different purposes. I see
no reason we can't have two.

Look, all I'm asking here is you try it. If things turn into a mess
(which I doubt), or another __cxa_demangle implementation can match
functionality, I'll revisit this issue. This seems sane. Until then, I
consider this issue closed, and thank you very much for your valued
participation.

>So, we've now duplicated that test.  And the old demangler (the one that
>will be used in binutils and GDB for the forseeable future) is not
>benefiting from the new tests.

... because they mostly fail. There are ~30 FAILs, and ~9 cores. 

The regression tests come mostly from GNATS. These PR's have been around
for months, close to and over a year in some cases, without any
movement. Perhaps you could take a look at them?

-benjamin


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]