gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Gordon Sadler]

This is copied to gcc and libstdc++ MLs.

On onlinedocs/libstdc++/configopts.html it says:
--enable-cstdio=LIB 

Select a target-specific I/O package. As of libstdc++-v3 snapshot 2.91,
the choices are 'libio' to specify the GNU I/O package (from glibc, the
GNU C library), or 'stdio' to use a generic "C" abstraction. The default
is 'stdio'. 

So most people take the default and so did I till yesterday. Now that
would tell me that libio doesn't get much testing...

linux 2.4.3ac12 glibc2.2.2 i686

For reference:
http://gcc.gnu.org/ml/gcc-testresults/2001-04/msg00437.html
http://gcc.gnu.org/ml/gcc-testresults/2001-04/msg00448.html

Only difference between these builds, 24 hours of minimal CVS activity
and --enable-cstdio=libio.

                === g++ Summary ===

		# of expected passes            6429
		# of unexpected failures        3
		# of expected failures          86
		# of untested testcases         9
		gcc-3.0/gcc/testsuite/../g++ version 3.0 20010422 (prerelease)
		

               === g++ Summary ===

	       # of expected passes            5564
	       # of unexpected failures        868
	       # of expected failures          86
	       # of untested testcases         9
	       gcc-3.0/gcc/testsuite/../g++ version 3.0 20010423 (prerelease)


Is the GNU I/O specific package really that much worse? Or has a lot of
code been make to work on the generic stdio without pushing to make it
work under libio?


Gordon Sadler
	       

	       

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Benjamin Kosnik]

Using libio is deprecated. Last time I checked there were linkage issues, 
and I haven't had the time to track them down yet.

> Is the GNU I/O specific package really that much worse? Or has a lot of
> code been make to work on the generic stdio without pushing to make it
> work under libio?

The latter. 

-benjamin

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Gabriel Dos Reis]

Benjamin Kosnik <bkoz@redhat.com> writes:

| Using libio is deprecated. Last time I checked there were linkage issues, 
| and I haven't had the time to track them down yet.

I guess C++ I/O on top of C I/O is a temporary solution.

-- Gaby

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Jason Merrill]

>>>>> "Benjamin" == Benjamin Kosnik <bkoz@redhat.com> writes:

> Using libio is deprecated.

"Deprecated" usually means "we want this to go away".  Is that really the
case with libio?  I thought the performance advantage was significant.

Jason

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on makecheck

[Benjamin Kosnik]

> "Deprecated" usually means "we want this to go away".  

Which is what I meant to say, strangely enough. 

> Is that really the
> case with libio?  I thought the performance advantage was significant.

Apparently not.

-benjamin

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Ulrich Drepper]

Benjamin Kosnik <bkoz@redhat.com> writes:

> > Is that really the
> > case with libio?  I thought the performance advantage was significant.
> 
> Apparently not.

I cannot believe this.  The main advantage of libio is that you don't
have to unbuffer the C library stdin, stdout, stderr streams to enable
correct synchronization.  Unless you conditionally drop this
requirement you'll have bad performance problems.

-- 
---------------.                          ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on makecheck

[Benjamin Kosnik]

> I cannot believe this.  The main advantage of libio is that you don't
> have to unbuffer the C library stdin, stdout, stderr streams to enable
> correct synchronization.  Unless you conditionally drop this
> requirement you'll have bad performance problems.

as in, this was the only performace degredation noted.

-benjamin

Re: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Gabriel Dos Reis]

Jason Merrill <jason_merrill@redhat.com> writes:

| >>>>> "Benjamin" == Benjamin Kosnik <bkoz@redhat.com> writes:
| 
| > Using libio is deprecated.
| 
| "Deprecated" usually means "we want this to go away".  Is that really the
| case with libio?  I thought the performance advantage was significant.

I cannot believe libio is really meant to go away.  My understanding
was that were some binary compatiblity issues which need to be sorted
out.  

-- Gaby

RE: gcc-3_0-branch cstdio=libio == libstdc++ doing miserable on make check

[Rob Taylor]

>
> Jason Merrill <jason_merrill@redhat.com> writes:
>
> | >>>>> "Benjamin" == Benjamin Kosnik <bkoz@redhat.com> writes:
> |
> | > Using libio is deprecated.
> |
> | "Deprecated" usually means "we want this to go away".  Is that really the
> | case with libio?  I thought the performance advantage was significant.
>
> I cannot believe libio is really meant to go away.  My understanding
> was that were some binary compatiblity issues which need to be sorted
> out.
>

also does this mean std::wcout/wcin/wcerr functionality has been implemented for
stdio?

Rob Taylor