Another libio patch

Another libio patch

[H.J. Lu]

Here is another libio patch.

-- 
H.J. Lu (hjl@gnu.ai.mit.edu)
---
Mon Nov 17 08:39:17 1997  H.J. Lu  (hjl@gnu.ai.mit.edu)

	* stdstrbufs.cc (DEF_STDFILE): Use STD_VTABLE.

Index: stdstrbufs.cc
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/stdstrbufs.cc,v
retrieving revision 1.1.1.2
diff -u -r1.1.1.2 stdstrbufs.cc
--- stdstrbufs.cc	1997/09/17 20:12:34	1.1.1.2
+++ stdstrbufs.cc	1997/11/17 20:38:04
@@ -58,7 +58,7 @@
 #define DEF_STDFILE(NAME, FD, CHAIN, FLAGS) \
   static _IO_lock_t _IO_stdfile_##FD##_lock = _IO_lock_initializer; \
   struct _IO_FILE_plus NAME \
-    = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), &_IO_file_jumps}
+    = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), STD_VTABLE}
 #else
 #define DEF_STDFILE(NAME, FD, CHAIN, FLAGS) \
   struct _IO_FILE_plus NAME = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), STD_VTABLE}

Re: Another libio patch

[Jeffrey A Law]

  In message < m0xXY3s-0004ecC@ocean.lucon.org >you write:
  > Here is another libio patch.
  > 
  > -- 
  > H.J. Lu (hjl@gnu.ai.mit.edu)
  > ---
  > Mon Nov 17 08:39:17 1997  H.J. Lu  (hjl@gnu.ai.mit.edu)
  > 
  > 	* stdstrbufs.cc (DEF_STDFILE): Use STD_VTABLE.
I have no idea what this patch does since you didn't make any attempt to
describe the problem it's trying to fix.  I will not install it unless I
hear from Ulrich.

Ulrich -- is this patch correct?

  > 
  > Index: stdstrbufs.cc
  > ===================================================================
  > RCS file: /home/work/cvs/gnu/egcs/libio/stdstrbufs.cc,v
  > retrieving revision 1.1.1.2
  > diff -u -r1.1.1.2 stdstrbufs.cc
  > --- stdstrbufs.cc	1997/09/17 20:12:34	1.1.1.2
  > +++ stdstrbufs.cc	1997/11/17 20:38:04
  > @@ -58,7 +58,7 @@
  >  #define DEF_STDFILE(NAME, FD, CHAIN, FLAGS) \
  >    static _IO_lock_t _IO_stdfile_##FD##_lock = _IO_lock_initializer; \
  >    struct _IO_FILE_plus NAME \
  > -    = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), &_IO_file_jumps}
  > +    = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), STD_VTABLE}
  >  #else
  >  #define DEF_STDFILE(NAME, FD, CHAIN, FLAGS) \
  >    struct _IO_FILE_plus NAME = {FILEBUF_LITERAL(CHAIN, FLAGS, FD), STD_VTAB
  > LE}

Re: Another libio patch

[H.J. Lu]

> 
> 
>   In message < m0xXY3s-0004ecC@ocean.lucon.org >you write:
>   > Here is another libio patch.
>   > 
>   > -- 
>   > H.J. Lu (hjl@gnu.ai.mit.edu)
>   > ---
>   > Mon Nov 17 08:39:17 1997  H.J. Lu  (hjl@gnu.ai.mit.edu)
>   > 
>   > 	* stdstrbufs.cc (DEF_STDFILE): Use STD_VTABLE.
> I have no idea what this patch does since you didn't make any attempt to
> describe the problem it's trying to fix.  I will not install it unless I

iostream should the C++ vtable. You can see STD_VTABLE is
used for another case in the same file.

> hear from Ulrich.
> 
> Ulrich -- is this patch correct?

That is one of the many problems with glibc 2 and egcs. The newer
glibc 2.1 breaks egcs due to the changes to libio in glibc 2.1.
I'd like to see glibc 2.1 install those libio private header files
and libio in egcs uses them if they are available. We may be
able to support libc 5, glibc 2.0 and glibc 2.1 in egcs.

H.J.

Re: Another libio patch

[H.J. Lu]

> 
> What needs to be done so that -lpthread is not needed on linux/glibc2?
>  I have applied the patch to /usr/include/libc-lock.h.  Rebuilding
> libstdc++ doesn't fix it - libstdc++.so still references two symbols
> defined in pthread.  Do I need to rebuild glibc?
> 

That should not happen. With my patch, those 2 symbols should
be weak and ld won't complain.

-- 
H.J. Lu (hjl@gnu.ai.mit.edu)

Re: Another libio patch

[Neal Becker]

What needs to be done so that -lpthread is not needed on linux/glibc2?
 I have applied the patch to /usr/include/libc-lock.h.  Rebuilding
libstdc++ doesn't fix it - libstdc++.so still references two symbols
defined in pthread.  Do I need to rebuild glibc?

Re: Another libio patch

[H.J. Lu]

> 
> hjl@lucon.org (H.J. Lu) writes:
> 
> > Please discard my previous libio patch. This patch should
> > fix both glibc 2 and libc 5. Please let me know the result
> > on libc 5.3.12 on linux/x86.
> 
> This looks much better.  Though
> 
> >  #ifdef weak_alias
> >  weak_alias (_IO_vsprintf, vsprintf)
> > +#else
> > +#ifdef __ELF__
> > +#pragma weak vsprintf = _IO_vsprintf
> > +#endif
> >  #endif
> 
> I don't think we should use the alias for other ELF systems (Solaris etc).
> I think this should read
> 
> 	#if defined __linux__ && defined __ELF__
> 

That is fine.

Thanks.


H.J.

Re: Another libio patch

[Ulrich Drepper]

hjl@lucon.org (H.J. Lu) writes:

> Please discard my previous libio patch. This patch should
> fix both glibc 2 and libc 5. Please let me know the result
> on libc 5.3.12 on linux/x86.

This looks much better.  Though

>  #ifdef weak_alias
>  weak_alias (_IO_vsprintf, vsprintf)
> +#else
> +#ifdef __ELF__
> +#pragma weak vsprintf = _IO_vsprintf
> +#endif
>  #endif

I don't think we should use the alias for other ELF systems (Solaris etc).
I think this should read

	#if defined __linux__ && defined __ELF__

-- Uli
---------------.      drepper@cygnus.com  ,-.   Rubensstrasse 5
Ulrich Drepper  \    ,-------------------'   \  76149 Karlsruhe/Germany
Cygnus Solutions `--' drepper@gnu.ai.mit.edu  `------------------------

Another libio patch

[H.J. Lu]

Hi,

Please discard my previous libio patch. This patch should
fix both glibc 2 and libc 5. Please let me know the result
on libc 5.3.12 on linux/x86.

Thanks.


-- 
H.J. Lu (hjl@gnu.ai.mit.edu)
---
Thu Oct  9 07:08:41 1997  H.J. Lu  (hjl@gnu.ai.mit.edu)

	* libio.h (_IO_LOCK_T): Handle glibc 2 when _IO_MTSAFE_IO is
	not defined.

	* filedoalloc.c (_IO_file_doallocate): Don't call
	_IO_cleanup_registration_needed if __linux__ is defined.

	* iofclose.c (fclose): Make it weak alias of _IO_fclose if
	__ELF__ is defined.

	* iovsprintf.c (vsprintf): Make it weak alias of _IO_vsprintf
	if __ELF__ is defined.

	* iovsscanf.c (vsscanf): Make it weak alias of _IO_vsscanf if
	__ELF__ is defined.

	* config/linuxlibc1.mt (MT_CFLAGS): Defined as -D_G_HAVE_MMAP.
	(IO_OBJECTS): Add filedoalloc.o fileops.o genops.o iofclose.o
	iovsprintf.o iovsscanf.o strops.o.

Index: libio/filedoalloc.c
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/filedoalloc.c,v
retrieving revision 1.1.1.2
diff -u -r1.1.1.2 filedoalloc.c
--- filedoalloc.c	1997/09/08 01:01:01	1.1.1.2
+++ filedoalloc.c	1997/10/09 19:47:37
@@ -75,7 +75,7 @@
   char *p;
   struct stat st;
 
-#ifndef _LIBC
+#if !defined(_LIBC) && !defined(__linux__)
   /* If _IO_cleanup_registration_needed is non-zero, we should call the
      function it points to.  This is to make sure _IO_cleanup gets called
      on exit.  We call it from _IO_file_doallocate, since that is likely
Index: libio/iofclose.c
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/iofclose.c,v
retrieving revision 1.1.1.2
diff -u -r1.1.1.2 iofclose.c
--- iofclose.c	1997/09/08 01:01:02	1.1.1.2
+++ iofclose.c	1997/10/09 19:02:33
@@ -55,4 +55,8 @@
 
 #ifdef weak_alias
 weak_alias (_IO_fclose, fclose)
+#else
+#ifdef __ELF__
+#pragma weak fclose = _IO_fclose
+#endif
 #endif
Index: libio/iovsprintf.c
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/iovsprintf.c,v
retrieving revision 1.1.1.2
diff -u -r1.1.1.2 iovsprintf.c
--- iovsprintf.c	1997/09/08 01:01:05	1.1.1.2
+++ iovsprintf.c	1997/10/09 19:03:52
@@ -54,4 +54,8 @@
 
 #ifdef weak_alias
 weak_alias (_IO_vsprintf, vsprintf)
+#else
+#ifdef __ELF__
+#pragma weak vsprintf = _IO_vsprintf
+#endif
 #endif
Index: libio/iovsscanf.c
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/iovsscanf.c,v
retrieving revision 1.1.1.2
diff -u -r1.1.1.2 iovsscanf.c
--- iovsscanf.c	1997/09/08 01:01:05	1.1.1.2
+++ iovsscanf.c	1997/10/09 19:03:29
@@ -51,4 +51,8 @@
 #ifdef weak_alias
 weak_alias (_IO_vsscanf, __vsscanf)
 weak_alias (_IO_vsscanf, vsscanf)
+#else
+#ifdef __ELF__
+#pragma weak vsscanf = _IO_vsscanf
+#endif
 #endif
Index: libio/libio.h
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/libio.h,v
retrieving revision 1.1.1.6
diff -u -r1.1.1.6 libio.h
--- libio.h	1997/10/09 17:31:37	1.1.1.6
+++ libio.h	1997/10/10 21:53:01
@@ -157,15 +157,17 @@
 # endif
 #else
 /* XXX This will go away as soon as comthread is finished.  */
-# ifdef __linux__
+# if defined __GLIBC__ && __GLIBC__ >= 2
+#  define _IO_LOCK_T void *
+# else
+#  ifdef __linux__
 struct _IO_lock_t {
   void *ptr;
   short int field1;
   short int field2;
 };
-#  define _IO_LOCK_T struct _IO_lock_t
-# else
-typedef void _IO_lock_t;
+#   define _IO_LOCK_T struct _IO_lock_t
+#  endif
 # endif
 #endif
 
Index: libio/config/linuxlibc1.mt
===================================================================
RCS file: /home/work/cvs/gnu/egcs/libio/config/linuxlibc1.mt,v
retrieving revision 1.1.1.3
diff -u -r1.1.1.3 linuxlibc1.mt
--- linuxlibc1.mt	1997/10/09 17:31:39	1.1.1.3
+++ linuxlibc1.mt	1997/10/09 19:38:44
@@ -9,7 +9,9 @@
 LIBIO_INCLUDE=
 
 # We have those in libc.a.
-IO_OBJECTS=iogetc.o ioputc.o iofeof.o ioferror.o
+IO_OBJECTS=iogetc.o ioputc.o iofeof.o ioferror.o \
+	filedoalloc.o fileops.o genops.o iofclose.o \
+	iovsprintf.o iovsscanf.o strops.o
 STDIO_WRAP_OBJECTS=
 OSPRIM_OBJECTS=
 STDIO_OBJECTS=
@@ -19,3 +21,6 @@
 	indstream.h iomanip.h iostream.h istream.h ostream.h \
 	parsestream.h pfstream.h procbuf.h stdiostream.h stream.h \
 	streambuf.h strfile.h strstream.h
+
+# A bad kludge
+MT_CFLAGS=-D_G_HAVE_MMAP

Another libio patch

[H.J. Lu]

I forgot this one in my libio patch.

Sorry for that.

-- 
H.J. Lu (hjl@gnu.ai.mit.edu)
--
--- /dev/null	Wed Dec 31 16:00:00 1969
+++ libio/config/libio.mt	Mon Aug 25 18:42:20 1997
@@ -0,0 +1,13 @@
+# Used by systems which implement stdio upon libio.
+# If libio in libc is not compatible with thunks, we
+# have to compile a few things ourselves.
+# FIXME: Use vfprintf.c and vfscanf.c from glibc.
+IO_OBJECTS=filedoalloc.o floatconv.o genops.o fileops.o iovfprintf.o \
+	   iovfscanf.o iopadn.o ioseekoff.o ioseekpos.o outfloat.o \
+	   strops.o iofclose.o iopopen.o \
+	   iofdopen.o iofflush.o iofopen.o iofputs.o \
+	   iofwrite.o ioputs.o iosetbuffer.o iosetvbuf.o \
+	   iovsprintf.o iovsscanf.o iovasprintf.o iovsnprintf.o
+		    
+XCFLAGS = $(LIBCFLAGS) -D_IO_USE_DTOA -D_IO_USE_LIBC
+XCXXFLAGS = $(LIBCXXFLAGS) -D_IO_USE_DTOA -D_IO_USE_LIBC
--- /dev/null	Wed Dec 31 16:00:00 1969
+++ libio/config/mtsafe.mt	Mon Aug 25 18:42:20 1997
@@ -0,0 +1,4 @@
+MTCFLAGS = -D_IO_MTSAFE_IO
+MTCXXFLAGS = -D_IO_MTSAFE_IO
+XCFLAGS = $(LIBCFLAGS) $(MTCFLAGS)
+XCXXFLAGS = $(LIBCXXFLAGS) $(MTCXXFLAGS)