Re: [Patch V4 06/10] EmulatorPkg: Fix XCODE5 lldb issues

["Michael D Kinney" <michael.d.kinney@intel.com>]

Jordan,

It is not a typo.

Andrew generated the XCODE specific changes, so they have 
been tested by him.  I have also reviewed and tested the XCODE
changes and verified that all 6 combinations build and boot
to shell (IA32/X64 for RELEASE/DEBUG/NOOPT). Since they are
all related to making EmulatorPkg work, we decided to fold
them into the same patch set that was already being reviewed.

I also verified build and boot to shell for 6 combinations
on GCC5 (IA32/X64 for RELEASE/DEBUG/NOOPT) and the 12
combinations of VS2015/VS2017, IA323/X64, RELEASE/DEBUG/NOOPT.

I have been working on some CI experiments using Azure Pipelines.
Here is a pointer to the build logs for all the combinations 
listed above.

https://dev.azure.com/mikekinney/edk2-ci/_build/results?buildId=312

Mike

> -----Original Message-----
> From: Justen, Jordan L
> Sent: Friday, August 16, 2019 12:41 AM
> To: Kinney, Michael D <michael.d.kinney@intel.com>;
> devel@edk2.groups.io
> Cc: Ni, Ray <ray.ni@intel.com>; Andrew Fish
> <afish@apple.com>
> Subject: Re: [Patch V4 06/10] EmulatorPkg: Fix XCODE5
> lldb issues
> 
> On 2019-08-15 19:14:33, Michael D Kinney wrote:
> > Fix scripts to support lldb symbolic debugging when
> using XCODE5 tool
> > chain.
> >
> > Cc: Jordan Justen <jordan.l.justen@intel.com>
> > Cc: Ray Ni <ray.ni@intel.com>
> > Cc: Michael D Kinney <michael.d.kinney@intel.com>
> > Signed-off-by: Andrew Fish <afish@apple.com>
> 
> Is this a Cc/Signed-off-by typo? (See also, patches 7-
> 10).
> 
> This makes me wonder if you are taking advantage of the
> git commit -s switch to add your Signed-off-by.
> 
> Also, I'm wondering if you are taking advantage of git-
> send-email automatically Cc'ing the addresses you
> listed in the commit message.
> (I thought it Cc'd for the author and Cc tags, but I
> wasn't sure about the Signed-off-by tag, and yet I see
> Andrew was Cc'd.)
> 
> There's a couple long lines below. You could use \ at
> the end of the line to split the .sh line. I think the
> cd can be a separate command in a shell script. (Not in
> make)
> 
> I hope someone that uses the XCODE toolchain could
> review/check the XCODE patches.
> 
> -Jordan
> 
> > ---
> >  EmulatorPkg/Unix/lldbefi.py |  8 +++++---
> >  EmulatorPkg/build.sh        | 17 ++---------------
> >  2 files changed, 7 insertions(+), 18 deletions(-)
> >
> > diff --git a/EmulatorPkg/Unix/lldbefi.py
> b/EmulatorPkg/Unix/lldbefi.py
> > index 218326b8cb..099192d8b5 100755
> > --- a/EmulatorPkg/Unix/lldbefi.py
> > +++ b/EmulatorPkg/Unix/lldbefi.py
> > @@ -346,6 +346,7 @@ def TypePrintFormating(debugger):
> >      debugger.HandleCommand("type summary add CHAR8 -
> -python-function lldbefi.CHAR8_TypeSummary")
> >      debugger.HandleCommand('type summary add --regex
> "CHAR8
> > \[[0-9]+\]" --python-function
> lldbefi.CHAR8_TypeSummary')
> >
> > +    debugger.HandleCommand('setting set frame-format
> "frame
> > + #${frame.index}: ${frame.pc}{
> > +
> ${module.file.basename}{:${function.name}()${function.p
> c-offset}}}{
> > + at ${line.file.fullpath}:${line.number}}\n"')
> >
> >  gEmulatorBreakWorkaroundNeeded = True
> >
> > @@ -381,15 +382,16 @@ def
> LoadEmulatorEfiSymbols(frame, bp_loc , internal_dict):
> >      Error = lldb.SBError()
> >      FileNamePtr = frame.FindVariable
> ("FileName").GetValueAsUnsigned()
> >      FileNameLen = frame.FindVariable
> > ("FileNameLength").GetValueAsUnsigned()
> > +
> >      FileName =
> frame.thread.process.ReadCStringFromMemory
> (FileNamePtr, FileNameLen, Error)
> >      if not Error.Success():
> >          print "!ReadCStringFromMemory() did not find
> a %d byte C string at %x" % (FileNameLen, FileNamePtr)
> >          # make breakpoint command contiue
> > -        frame.GetThread().GetProcess().Continue()
> > +        return False
> >
> >      debugger = frame.thread.process.target.debugger
> >      if frame.FindVariable
> ("AddSymbolFlag").GetValueAsUnsigned() == 1:
> > -        LoadAddress = frame.FindVariable
> ("LoadAddress").GetValueAsUnsigned()
> > +        LoadAddress = frame.FindVariable
> > + ("LoadAddress").GetValueAsUnsigned() - 0x240
> >
> >          debugger.HandleCommand ("target modules add
> %s" % FileName)
> >          print "target modules load --slid 0x%x %s" %
> (LoadAddress,
> > FileName) @@ -405,7 +407,7 @@ def
> LoadEmulatorEfiSymbols(frame, bp_loc , internal_dict):
> >                      print "!lldb.target.RemoveModule
> (%s) FAILED" %
> > SBModule
> >
> >      # make breakpoint command contiue
> > -    frame.thread.process.Continue()
> > +    return False
> >
> >  def GuidToCStructStr (guid, Name=False):
> >    #
> > diff --git a/EmulatorPkg/build.sh
> b/EmulatorPkg/build.sh index
> > 60056e1b6c..35912a7775 100755
> > --- a/EmulatorPkg/build.sh
> > +++ b/EmulatorPkg/build.sh
> > @@ -209,21 +209,8 @@ fi
> >  if [[ "$RUN_EMULATOR" == "yes" ]]; then
> >    case `uname` in
> >      Darwin*)
> > -      #
> > -      # On Darwin we can't use dlopen, so we have to
> load the real PE/COFF images.
> > -      # This .gdbinit script sets a breakpoint that
> loads symbols for the PE/COFFEE
> > -      # images that get loaded in Host
> > -      #
> > -      if [[ "$CLANG_VER" == *-ccc-host-triple* ]]
> > -      then
> > -      # only older versions of Xcode support -ccc-
> host-tripe, for newer versions
> > -      # it is -target
> > -        cp $WORKSPACE/EmulatorPkg/Unix/lldbefi.py
> "$BUILD_OUTPUT_DIR/${BUILDTARGET}_$TARGET_TOOLS/$PROCES
> SOR"
> > -        cd $BUILD_ROOT_ARCH; /usr/bin/lldb --source
> $WORKSPACE/EmulatorPkg/Unix/lldbinit Host
> > -        exit $?
> > -      else
> > -        cp $WORKSPACE/EmulatorPkg/Unix/.gdbinit
> "$BUILD_OUTPUT_DIR/${BUILDTARGET}_$TARGET_TOOLS/$PROCES
> SOR"
> > -      fi
> > +      cd $BUILD_ROOT_ARCH; /usr/bin/lldb -o "command
> script import $WORKSPACE/EmulatorPkg/Unix/lldbefi.py" -
> o 'script lldb.debugger.SetAsync(True)' -o "run" ./Host
> > +      exit $?
> >        ;;
> >    esac
> >
> > --
> > 2.21.0.windows.1
> >