Re: GDB 9.0.90 available for testing

[Eli Zaretskii <eliz@gnu.org>]

> From: Joel Brobecker <brobecker@adacore.com>
> Date: Wed, 11 Dec 2019 22:47:45 +0100 (CET)
> 
> 
>     ftp://sourceware.org/pub/gdb/snapshots/branch/gdb-9.0.90.tar.xz
> 
> A gzip'ed version is also available: gdb-9.0.90.tar.gz.
> 
> Please give it a test if you can and report any problems you might find.

I've built this with mingw.org's MinGW, using GCC 8.3.0 and Binutils
2.33, and found the following issues:

1. The produced gdb.exe and gdbserver.exe depend on the libstdc++ and
   libgcc DLLs.  I thought this was supposed to be prevented, or did I
   dream?  Should I force -static-libgcc etc. switches via
   configure-time variables?

2. readline/colors.c fails to compile because it uses the likes of
   S_IXGRP and S_IXOTH, which aren't defined in MinGW.  The solution
   is to patch readline/posixstat.h to add the missing defines (it
   tries to do so, but makes assumptions that don't do a perfect job).

3. When configure runs in gdb/, it displays a warning thusly:

     checking for pthread-config... no
     /d/usr/eli/utils/gdb-9.0.90/gdb/configure: line 14342: test: =: unary operator expected

   It is triggered by this line:

     if test $gdb_cv_cxx_std_thread = yes; then

   but I have no idea what's wrong with this.  Any hints?

4. libctf fails to compile due to 3 compilation errors.  I already
   found these problems when building Binutils 2.33 and reported them
   on Bugzilla (https://sourceware.org/bugzilla/show_bug.cgi?id=25155),
   but it seems nothing was done about that in the meantime.

5. A compilation warning in gdb/, which wasn't there in GDB 8.3:

       CXX    record-btrace.o
     In file included from ../../gdb-9.0.90/gdb/exceptions.h:23,
		      from ../../gdb-9.0.90/gdb/utils.h:24,
		      from ../../gdb-9.0.90/gdb/defs.h:652,
		      from ../../gdb-9.0.90/gdb/record-btrace.c:22:
     ../../gdb-9.0.90/gdb/ui-out.h: In function 'void btrace_insn_history(ui_out*, const btrace_thread_info*, const btrace_insn_iterator*, const btrace_insn_iterator*, gdb_disassembly_flags)':
     ../../gdb-9.0.90/gdb/ui-out.h:349:18: warning: 'asm_list.ui_out_emit_type<(ui_out_type)1>::m_uiout' may be used uninitialized in this function [-Wmaybe-uninitialized]
	  m_uiout->end (Type);
	  ~~~~~~~~~~~~~^~~~~~
     ../../gdb-9.0.90/gdb/record-btrace.c:779:35: note: 'asm_list.ui_out_emit_type<(ui_out_type)1>::m_uiout' was declared here
	gdb::optional<ui_out_emit_list> asm_list;
					^~~~~~~~

   Any suggestions how to fix this?

6. One last comment about "gdb -tui": while stepping with "next"
   through a program (in this case, GDB's only binary), sometimes the
   source window becomes blank and shows only "No source" in it.  This
   happens when the command windo shows some DWARF-related error
   message.  Is this expected?  I don't think I saw this in previous
   versions of GDB.

Please tell me how to go about fixing these problems on the release
branch.  Do we still maintain our separate copy of Readline, or do I
need to report to its upstream maintainer?  And what to do about
libctf? how can I speed up the handling of those problems upstream?

Thanks.