FYI: fix dumpbin symbol pipe (was: state of MSVC patches for HEAD)

[Ralf Wildenhues]

Hi Peter,

* Peter Ekberg wrote on Wed, Aug 31, 2005 at 05:52:43PM CEST:
> I wrote:
> > Ralf Wildenhues wrote:
> > > * Peter Ekberg wrote on Tue, Aug 30, 2005 at 12:43:16PM CEST:
> > > > 
> > > awk is a possible candidate, but please try to stay within POSIX
> > > features.
> > 
> > There is *little* documentation that I have found which is helpful.
> > 
> > Here is some:
> > http://msdn.microsoft.com/library/en-us/vccore/html/_dumpbin_2f.symbols.asp

Thanks.

> > I look at the output and then try to make educated guesses,
> > which has not been 100% successful. I think I'm closing in
> > on the target though.
*snip*
> > In the above there are five section lines, and SECT5 is the
> > Highest SECTn there is. This is documented in the above URL.
> > 
> > The following are my educated guesses:
> > 
> > Section lines are never repeated.
> > 
> > Immediately before the nth section line, a "section symbol"
> > appears which is the first reference with SECTn. This symbol
> > appears to be the section name, so may contain info as
> > to what kind of segment it is. I mean .drectve, .debug$S,
> > .data, .text and .rdata are probably pretty standard names.
> > To me, the last three look familiar, but using the section
> > name to deduce section type seems fragile, so I'd go for
> > the "(pick any)" string in the section line following the
> > first "section symbol" of the new section instead.
> > 
> > So, I think what's needed is to keep track of if which
> > sections are marked "(pick any)", and ignore any symbols
> > from those sections.
> > 
> > This will be cumbersome to do with sed as symbols from e.g.
> > the 4th section may appear after the 5th section header
> > (see _foo1_LTX_hello, which should be exported), so it is
> > best to keep track of all previous sections.
> > 
> > I think awk will be a good tool to solve this, but awk is
> > another tool which I will need to read up on...
> 
> Here's an attempt,I bet it could be made more efficient by
> someone with more awk experience...
> 
> I have tried to stay away from gawk-specific constructs,
> as documented by "info gawk", but allowed myself to
> use posix stuff (split with regexp in 3rd arg), even if not
> all awk implementations are posix compliant. But who does
> not run gawk on MinGW and Cygwin?

OK, I think we can live with this for now.

> M4 got in the way a bit, so there are a couple of $][0 and
> a $][3 in there that makes it hard(er) to read...
> 
> Also, the echo of the command is warped during build, which
> is a bit unfortunate, e.g:

*snip*
> As seen, $0 and $3 are replaced by "./libtool" and ""
> respectively. I guess $][0 could be replaced by $""0 to
> get around that, but that's ugly^Hier...

I fixed all of this the same way as in the AIX case: change 
`$][3' to `$ 3'.  awk does not mind the space.

> Works for me (MSVC 6 and now also MSVC 2003), ok to commit?

Works much better for me than anything before that, so I committed the
patch below to HEAD.

Thanks a lot for your work!

A little note to keep in mind: before we release, we have to add a
statement that all of the new (even if documented) features are
experimental and bound to change.  (I'm not sure if going back to not
using $NM as dumpbin would be a better idea, and I want to keep this
option available to our discretion.)

Cheers,
Ralf

2005-09-08  Peter Ekberg  <address@hidden>

        * libltdl/m4/libtool.m4 (_LT_CMD_GLOBAL_SYMBOLS) [dumpbin]:
        Symbols in "pick any" sections are not global. Use
        awk, not sed, in the global_symbol_pipe so that keeping
        track of sections is easier.
        * doc/libtool.texi (libtool script contents): Update to hint
        at new findings concerning "pick any" sections.

Index: doc/libtool.texi
===================================================================
RCS file: /cvsroot/libtool/libtool/doc/libtool.texi,v
retrieving revision 1.197
diff -u -r1.197 libtool.texi
--- doc/libtool.texi    1 Sep 2005 11:54:50 -0000       1.197
+++ doc/libtool.texi    8 Sep 2005 08:18:33 -0000
@@ -5017,7 +5017,8 @@
 @end example
 
 The @var{size} of the global variables are not zero and the @var{section}
-of the global functions are not "UNDEF".
+of the global functions are not "UNDEF". Symbols in "pick any" sections
+("pick any" appears in the section header) are not global either.
 @end defvar
 
 @defvar RANLIB
Index: libltdl/m4/libtool.m4
===================================================================
RCS file: /cvsroot/libtool/libtool/libltdl/m4/libtool.m4,v
retrieving revision 1.16
diff -u -r1.16 libtool.m4
--- libltdl/m4/libtool.m4       8 Sep 2005 07:00:03 -0000       1.16
+++ libltdl/m4/libtool.m4       8 Sep 2005 08:18:34 -0000
@@ -3078,11 +3078,17 @@
     # and D for any global variable.
     # Also find C++ and __fastcall symbols from MSVC++,
     # which start with @ or ?.
-    lt_cv_sys_global_symbol_pipe="$SED -n -e ['/ UNDEF [^|]*()/d; / 00* UNDEF 
/d;
-       s/.*().*External *| *$ac_symprfx$sympat.*/T $ac_symprfx\1 \1/p;
-       s/.*External *| *$ac_symprfx$sympat.*/D $ac_symprfx\1 \1/p;
-       s/.*().*External *| *\(address@hidden@?]*\).*/T \1 \1/p;
-       s/.*External *| *\(address@hidden@?]*\).*/D \1 \1/p']"
+    lt_cv_sys_global_symbol_pipe="$AWK ['
+      {last_section=section; section=\$ 3}
+      /Section length .*#relocs.*(pick any)/{hide[last_section]=1}
+      \$ 0!~/External *\|/{next}
+      / 0+ UNDEF /{next}; / UNDEF \([^|]\)*()/{next}
+      {if(hide[section]) next}
+      {f=0}; \$ 0~/\(\).*\|/{f=1}; {printf f ? \"T \" : \"D \"}
+      {split(\$ 0, a, /\||\r/); split(a[2], s)}
+      s[1]~/address@hidden/{print s[1], s[1]; next}
+      s[1]~prfx {split(s[1],t,\"@\"); print t[1], substr(t[1],length(prfx))}
+      ' prfx=^$ac_symprfx]"
   else
     lt_cv_sys_global_symbol_pipe="sed -n -e 's/^.*[[   
]]\($symcode$symcode*\)[[       ]][[    
]]*$ac_symprfx$sympat$opt_cr$/$symxfrm/p'"
   fi