[cfe-dev] Missing debug information for Windows

[Dennis Luehring]

>I'm just pointing out that I seem to
>have no "collegues" around there,
>working like me

not 100% true :)

my pros:
-found serveral small bugs with cl-clang diagnostics
-helped porting Windows sources to Linux
-nearly out of the box compilation

my cons:
-no proper VS-Debug support
-not-working exceptions
-partialy compilation problems with throw/catch using code
-compilation speed is slower then cl - even with disabled multi-thread 
compile and /MP
-sometimes headers are not found

beside my cons im very suprised about the quality of cl-clang working 
and results

Am 21.06.2015 um 09:52 schrieb Javier Múgica:
> Thank you for your information. It seems there are still many open issues for me to be able to use Clang on a daily basis on Windows.
>
> I suspect there aren't many Clang users targeting the Visual Sutio libraries. You cannot currently successfully compile a project in Clang if you include tgmaht.h, not even with VS2015 (because of a pitfall in VS, no in Clang) and my propossed patch to that file sent to cfe-commits dind't raise any reply. I don't care about it since it is easy to work around. I'm just pointing out that I seem to have no "collegues" around there, working like me, trying to use C11, VS and Clang.
>
> Regards,
>
> Javier A. Múgica
>
> On Fri, 19 Jun 2015 10:32:22 -0700, David Blaikie <dblaikie at gmail.com> wrote:
>
> > On Fri, Jun 19, 2015 at 12:22 AM, Javier Múgica <javier_3 at runbox.com> wrote:
> >
> > > Hello:
> > >
> > > I am facing a critic problem with the compiler, which I hope will have an
> > > easy solution. I am using it on Windows and use Visual Studio to debug the
> > > applications. The information of which source lines correspond to which
> > > machine instructions is right and I can break into the code, execute one
> > > instruction at a time, etc. But all the information regarding the variables
> > > is missing, so I cannot know the values they store if not by guessing upon
> > > the assembly code where is stored each variable and asking VS to show me
> > > the value of, say, **(double**)(esp+01Ch).
> > >
> > > I have tried both clang and clang-cl, and both -O0 and -O1, and both VS
> > > 2010 and VS 2015. I use clang with the -c option (do not link), and then
> > > call the VS linker, which is what clang does if the -c option is not used.
> > >
> > > The Microsoft compiler generates an intermediate .pdb file when compiling
> > > and assemblying, named vc100.pdb (the name is of course irrelevant), and
> > > afterwards, when linking, the definite pdb file, foo.pdb for instance. When
> > > using clang the intermediate file is not generated.  May that be the reason
> > > or has it nothing to do with my problem?
> > >
> >
> > That's the reason, but there is no easy solution - Clang does not generate
> > PDB files at the moment. The debug story on windows is still uncertain and
> > a fair way out - it may entail integrating something like LLDB into MSVC to
> > use the existing DWARF debug info Clang can produce.
> >
> > For now you could use LLDB directly on windows, though I think support
> > there is still work-in-progress too.
> >
> > Sorry,
> >
> > - David
> >
> >
> > >
> > > Lest's hope you can help me, otherwise I will have to stop using clang
> > > just when I was beginning with it.
> > >
> > > -- Javier
> > >
> > > _______________________________________________
> > > cfe-dev mailing list
> > > cfe-dev at cs.uiuc.edu
> > > http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
> > >
>
>
>
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev