Re: Mainline Is Busted for Java

Re: Mainline Is Busted for Java

[John David Anglin]

Java also broke yesterday on hppa-linux:

/home/dave/gcc-3.5/objdir/gcc/xgcc -shared-libgcc -B/home/dave/gcc-3.5/objdir/gc
c/ -nostdinc++ -L/home/dave/gcc-3.5/objdir/hppa-linux/libstdc++-v3/src -L/home/d
ave/gcc-3.5/objdir/hppa-linux/libstdc++-v3/src/.libs -B/home/dave/opt/gnu/gcc/gc
c-3.5/hppa-linux/bin/ -B/home/dave/opt/gnu/gcc/gcc-3.5/hppa-linux/lib/ -isystem
/home/dave/opt/gnu/gcc/gcc-3.5/hppa-linux/include -isystem /home/dave/opt/gnu/gc
c/gcc-3.5/hppa-linux/sys-include -DHAVE_CONFIG_H -I. -I../../../gcc/libjava -I./
include -I./gcj -I../../../gcc/libjava -Iinclude -I../../../gcc/libjava/include
-I/home/dave/gcc-3.5/gcc/boehm-gc/include -DPACKAGE_NAME= -DPACKAGE_TARNAME= -DP
ACKAGE_VERSION= -DPACKAGE_STRING= -DPACKAGE_BUGREPORT= -DGC_LINUX_THREADS=1 -D_R
EENTRANT=1 -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_STDL
IB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=
1 -DHAVE_STDINT_H=1 -DHAVE_UNISTD_H=1 -DSILENT=1 -DNO_SIGNALS=1 -DALL_INTERIOR_P
OINTERS=1 -DJAVA_FINALIZATION=1 -DGC_GCJ_SUPPORT=1 -DATOMIC_UNCOLLECTABLE=1 -I..
/../../gcc/libjava/libltdl -I../../../gcc/libjava/libltdl -I../../../gcc/libjava
/.././libjava/../gcc -I../../../gcc/libjava/../zlib -I../../../gcc/libjava/../li
bffi/include -I../libffi/include -O2 -g -O2 -fno-rtti -fnon-call-exceptions -fdo
llars-in-identifiers -Wswitch-enum -I/usr/X11R6/include -W -Wall -D_GNU_SOURCE -
DPREFIX=\"/home/dave/opt/gnu/gcc/gcc-3.5\" -DLIBDIR=\"/home/dave/opt/gnu/gcc/gcc
-3.5/lib\" -DBOOT_CLASS_PATH=\"/home/dave/opt/gnu/gcc/gcc-3.5/share/java/libgcj-
3.5.0.jar\" -g -O2 -D_GNU_SOURCE -MD -MT java/net/natInetAddress.lo -MF java/net
/natInetAddress.pp -c java/net/natInetAddress.cc  -fPIC -DPIC -o java/net/.libs/
natInetAddress.o
java/net/natInetAddress.cc: In static member function `static JArray<java::net::InetAddress*>* java::net::InetAddress::lookup(java::lang::String*, java::net::InetAddress*, jboolean)':
java/net/natInetAddress.cc:116: error: aggregate `hostent_data fixed_buffer' has incomplete type and cannot be defined
java/net/natInetAddress.cc:148: error: cannot convert `hostent_data*' to `char*' for argument `3' to `int gethostbyname_r(const char*, hostent*, char*, size_t, hostent**, int*)'
java/net/natInetAddress.cc:201: error: cannot convert `hostent_data*' to `char*' for argument `5' to `int gethostbyaddr_r(const void*, __socklen_t, int, hostent*, char*, size_t, hostent**, int*)'

For some reason, this error didn't stop the build.  I guess that's the problem
with the gcc driver that's been reported by Kenner and others.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: Mainline Is Busted for Java

[Caroline Tice]

I am very sorry that my driver patch appears to have broken Java; I am 
working on a fix which will hopefully
be ready in an hour (or possibly less).  I had not gotten around to 
testing the original patch on Java, but it was
on my list of things I had planned to do yesterday anyway;  
unfortunately other people found that it broke Java before then.  I 
apologize for not testing on Java previously.  I won't do this again.  
I repeat, I am very sorry.

-- Caroline Tice
ctice@apple.com

On Apr 7, 2004, at 7:07 AM, John David Anglin wrote:

> Java also broke yesterday on hppa-linux:
>
> /home/dave/gcc-3.5/objdir/gcc/xgcc -shared-libgcc 
> -B/home/dave/gcc-3.5/objdir/gc
> c/ -nostdinc++ -L/home/dave/gcc-3.5/objdir/hppa-linux/libstdc++-v3/src 
> -L/home/d
> ave/gcc-3.5/objdir/hppa-linux/libstdc++-v3/src/.libs 
> -B/home/dave/opt/gnu/gcc/gc
> c-3.5/hppa-linux/bin/ -B/home/dave/opt/gnu/gcc/gcc-3.5/hppa-linux/lib/ 
> -isystem
> /home/dave/opt/gnu/gcc/gcc-3.5/hppa-linux/include -isystem 
> /home/dave/opt/gnu/gc
> c/gcc-3.5/hppa-linux/sys-include -DHAVE_CONFIG_H -I. 
> -I../../../gcc/libjava -I./
> include -I./gcj -I../../../gcc/libjava -Iinclude 
> -I../../../gcc/libjava/include
> -I/home/dave/gcc-3.5/gcc/boehm-gc/include -DPACKAGE_NAME= 
> -DPACKAGE_TARNAME= -DP
> ACKAGE_VERSION= -DPACKAGE_STRING= -DPACKAGE_BUGREPORT= 
> -DGC_LINUX_THREADS=1 -D_R
> EENTRANT=1 -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 
> -DHAVE_STDL
> IB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 
> -DHAVE_INTTYPES_H=
> 1 -DHAVE_STDINT_H=1 -DHAVE_UNISTD_H=1 -DSILENT=1 -DNO_SIGNALS=1 
> -DALL_INTERIOR_P
> OINTERS=1 -DJAVA_FINALIZATION=1 -DGC_GCJ_SUPPORT=1 
> -DATOMIC_UNCOLLECTABLE=1 -I..
> /../../gcc/libjava/libltdl -I../../../gcc/libjava/libltdl 
> -I../../../gcc/libjava
> /.././libjava/../gcc -I../../../gcc/libjava/../zlib 
> -I../../../gcc/libjava/../li
> bffi/include -I../libffi/include -O2 -g -O2 -fno-rtti 
> -fnon-call-exceptions -fdo
> llars-in-identifiers -Wswitch-enum -I/usr/X11R6/include -W -Wall 
> -D_GNU_SOURCE -
> DPREFIX=\"/home/dave/opt/gnu/gcc/gcc-3.5\" 
> -DLIBDIR=\"/home/dave/opt/gnu/gcc/gcc
> -3.5/lib\" 
> -DBOOT_CLASS_PATH=\"/home/dave/opt/gnu/gcc/gcc-3.5/share/java/libgcj-
> 3.5.0.jar\" -g -O2 -D_GNU_SOURCE -MD -MT java/net/natInetAddress.lo 
> -MF java/net
> /natInetAddress.pp -c java/net/natInetAddress.cc  -fPIC -DPIC -o 
> java/net/.libs/
> natInetAddress.o
> java/net/natInetAddress.cc: In static member function `static 
> JArray<java::net::InetAddress*>* 
> java::net::InetAddress::lookup(java::lang::String*, 
> java::net::InetAddress*, jboolean)':
> java/net/natInetAddress.cc:116: error: aggregate `hostent_data 
> fixed_buffer' has incomplete type and cannot be defined
> java/net/natInetAddress.cc:148: error: cannot convert `hostent_data*' 
> to `char*' for argument `3' to `int gethostbyname_r(const char*, 
> hostent*, char*, size_t, hostent**, int*)'
> java/net/natInetAddress.cc:201: error: cannot convert `hostent_data*' 
> to `char*' for argument `5' to `int gethostbyaddr_r(const void*, 
> __socklen_t, int, hostent*, char*, size_t, hostent**, int*)'
>
> For some reason, this error didn't stop the build.  I guess that's the 
> problem
> with the gcc driver that's been reported by Kenner and others.
>
> Dave
> -- 
> J. David Anglin                                  
> dave.anglin@nrc-cnrc.gc.ca
> National Research Council of Canada              (613) 990-0752 (FAX: 
> 952-6602)
>

Re: Mainline Is Busted for Java

[Ranjit Mathew]

Mark Wielaard wrote:
> But note that this bug somehow makes configure pick up the installed gcj
> and gcjh and not the just build gcj and gcjh. We are seeing the breakage
> because on our systems the installed gcj and gcjh have bugs that show up
> against the current libgcj implementation. What might have happened is
> that she had a recently build gcj in here PATH that doesn't trigger the
> bugs.

Thank you for explaining that!

I was trying so hard to figure out exactly why
these changes would cause the errors I was
seeing in GCJ...

I have uninstalled the vendor-installed GCJ (v3.2.3).

Thanks,
Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: Mainline Is Busted for Java

[Ian Lance Taylor]

Mark Wielaard <mark@klomp.org> writes:

> P.S. We are talking among ourselves about this issue here and on irc.
> Has somebody actually contacted her about what we are seeing?

Caroline has already sent a patch to gcc-patches.  Somebody needs to
approve it.

Ian

Re: Mainline Is Busted for Java

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 1133 bytes --]

Hi,

On Wed, 2004-04-07 at 14:34, Andrew Pinski wrote:
> On Apr 7, 2004, at 08:26, Ranjit Mathew wrote:
> 
> > Andrew Haley wrote:
> >>> Andrew Pinski thinks this is because of Caroline Tice's
> >>> recent patch:
> >>
> >> This is weird.  I mean, how come the patch was checked in if the build
> >> failed.  Anyway, I recommend we give Caroline a little while to fix
> >> it, if indeed it is caused by her patch.
> 
> Because it looks like she did not test by building libjava which is 
> enabled
> by default on PPC Darwin by the way, so there should be no excuse for 
> this
> one.

But note that this bug somehow makes configure pick up the installed gcj
and gcjh and not the just build gcj and gcjh. We are seeing the breakage
because on our systems the installed gcj and gcjh have bugs that show up
against the current libgcj implementation. What might have happened is
that she had a recently build gcj in here PATH that doesn't trigger the
bugs.

Cheers,

Mark

P.S. We are talking among ourselves about this issue here and on irc.
Has somebody actually contacted her about what we are seeing?

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Mainline is Busted for Java

[Nathanael Nerode]

>PS: "make new-bootstrap" has been producing a gcj that
>fails spectacularly on i686-pc-linux-gnu. What *is* this
>make target for?

It's the experimental target intended to (eventually) replace the
current "bootstrap" logic.

Of course, it can't do that until it *works*.  Currently I know that
it breaks libada pretty badly, and I'm working on fixing that on libada-branch
(but my fixes are breaking libada in other ways, so I have to find and fix
those first...)

I haven't seen problems with GCJ which don't occur with ordinary bootstrap.
Could you please tell me what "spectacular" failures you're seeing, so I
can fix them? :-)

-- 
Make sure your vote will count.
http://www.verifiedvoting.org/

Re: Mainline Is Busted for Java

[Ranjit Mathew]

Andrew Pinski wrote:
> Because it looks like she did not test by building libjava which is 
> enabled
> by default on PPC Darwin by the way, so there should be no excuse for 
> this
> one.

However:

  http://gcc.gnu.org/ml/gcc-patches/2004-03/msg02327.html

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: Mainline Is Busted for Java

[Andrew Pinski]

On Apr 7, 2004, at 08:26, Ranjit Mathew wrote:

> Andrew Haley wrote:
>>> Andrew Pinski thinks this is because of Caroline Tice's
>>> recent patch:
>>
>> This is weird.  I mean, how come the patch was checked in if the build
>> failed.  Anyway, I recommend we give Caroline a little while to fix
>> it, if indeed it is caused by her patch.

Because it looks like she did not test by building libjava which is 
enabled
by default on PPC Darwin by the way, so there should be no excuse for 
this
one.


> FWIW, I just successfully bootstrapped (c,c++,java) on
> i686-pc-linux-gnu by reverting Caroline's patch:
>
>   http://gcc.gnu.org/ml/gcc-patches/2004-03/msg02101.html
>
> in my local tree.
>
> So Andrew (Pinski) seems to be correct in his conjecture.

That is because I finally reverted the patch also and did a bootstrap
my self when I figured out the problem.

Also note the problem is being tracked by PR 14874.


> Ranjit.
>
> PS: "make new-bootstrap" has been producing a gcj that
> fails spectacularly on i686-pc-linux-gnu. What *is* this
> make target for?


new-bootstrap is the new top level bootstrap target in the
Makefiles, it is experimental at this point.

Thanks,
Andrew Pinski

Re: Mainline Is Busted for Java

[Ranjit Mathew]

Andrew Haley wrote:
>  > Andrew Pinski thinks this is because of Caroline Tice's
>  > recent patch:
> 
> This is weird.  I mean, how come the patch was checked in if the build
> failed.  Anyway, I recommend we give Caroline a little while to fix
> it, if indeed it is caused by her patch.

FWIW, I just successfully bootstrapped (c,c++,java) on
i686-pc-linux-gnu by reverting Caroline's patch:

  http://gcc.gnu.org/ml/gcc-patches/2004-03/msg02101.html

in my local tree.

So Andrew (Pinski) seems to be correct in his conjecture.

Ranjit.

PS: "make new-bootstrap" has been producing a gcj that
fails spectacularly on i686-pc-linux-gnu. What *is* this
make target for?

-- 
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Mainline Is Busted for Java

[Andrew Haley]

Ranjit Mathew writes:
 > Hi,
 > 
 >   A bootstrap of mainline GCC fails for Java with:
 > ----------------------------- 8< -----------------------------
 > gcj -B/home/ranmath/src/gcc/build/i686-pc-linux-gnu/libjava/ --encoding=UTF-8 -Wno-deprecated -C -g -classpath '' -bootclasspath /home/ranmath/src/gcc/build/i686-pc-linux-gnu/libjava:/home/ranmath/src/gcc/gcc-20040407/libjava -d /home/ranmath/src/gcc/build/i686-pc-linux-gnu/libjava /home/ranmath/src/gcc/gcc-20040407/libjava/gnu/awt/LightweightRedirector.java
 > /home/ranmath/src/gcc/gcc-20040407/libjava/java/awt/font/TextAttribute.java:44:
 > Superclass `AttributedCharacterIterator.Attribute' of class `java.awt.font.TextAttribute' not found.
 >    public final class TextAttribute extends AttributedCharacterIterator.Attribute
 >                                             ^
 > 1 error
 > ----------------------------- 8< -----------------------------
 > 
 > Andrew Pinski thinks this is because of Caroline Tice's
 > recent patch:

This is weird.  I mean, how come the patch was checked in if the build
failed.  Anyway, I recommend we give Caroline a little while to fix
it, if indeed it is caused by her patch.

Andrew.