bad MELT tentative release 0.9.6 - help appreciated

[Basile Starynkevitch]

Hello All

I almost released MELT plugin 0.9.6 for GCC 4.6 & 4.7 but thanks to Emmanuel Haucourt I found a bug
which I cannot easily reproduce. Perhaps it is related to the differences (versions of make and of gcc)
between our machines (Ubuntu vs Debian/Sid)

In case you want to help and test it, you could get that thing from

http://gcc-melt.org/badmelt-r190008-0.9.6-plugin-for-gcc-4.6-or-4.7.tar.gz

the md5sum of that thing of 5606772 bytes is e636e1f7c4bb00b5f6c4a9d5b046167d
it is extracted from MELT branch svn 190008

(that URL is obviously temporary; I'll remove the thing when debugged the issue)

What fails on one machine but works on another is the following human procedure

Retrieve that tarball above

untar it (it will produce a melt-0.9.6-plugin-for-gcc-4.6-or-4.7 directory without any badmelt in it)

build the melt plugin as usual:

make
make install DESTDIR=/tmp/meltinst

Then install it e.g. with
sudo cp -v -r -p /tmp/meltinst/usr/. /usr/.

Then get the melt-examples with git, e.g. with
git clone https://github.com/bstarynk/melt-examples.git
Then (a hello world example)
cd melt-examples/ex01
make all
make run

Please report (e.g. on this gcc-melt@ list) any success or any failure.
In all cases, give the versions of GCC, of GNU make, of Bash you got
Please also give any relevant information useful to debug the issue

(a possible issue is that MELT is using random numbers as hash codes, so the generated C code varies from time to time)

Thanks.
--
Basile STARYNKEVITCH http://starynkevitch.net/Basile/
email: basile<at>starynkevitch<dot>net mobile: +33 6 8501 2359
8, rue de la Faiencerie, 92340 Bourg La Reine, France
*** opinions {are only mines, sont seulement les miennes} ***

[Pierre Vittet]

Hello,

I have quickly compiled melt following your instructions (no problem for
compiling/installing MELT).


I tried to make and run ex01 and I guess I could reproduce the bug (or
at least it produced one). I obtained a "The bug is not reproducible, so
it is likely a hardware or OS problem". See the log.

Pierre Vittet

----------------------------

dpierre@Birdy ~ %gcc -v
Utilisation des specs internes.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.6/lto-wrapper
Target: x86_64-linux-gnu
Configuré avec: ../src/configure -v --with-pkgversion='Debian 4.6.3-1'
--with-bugurl=file:///usr/share/doc/gcc-4.6/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++,go --prefix=/usr
--program-suffix=-4.6 --enable-shared --enable-linker-build-id
--with-system-zlib --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.6
--libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin
--enable-objc-gc --with-arch-32=i586 --with-tune=generic
--enable-checking=release --build=x86_64-linux-gnu
--host=x86_64-linux-gnu --target=x86_64-linux-gnu
Modèle de thread: posix
gcc version 4.6.3 (Debian 4.6.3-1)


----------------------------
dpierre@Birdy ex01 %make -v
GNU Make 3.81
...

---------------------------
dpierre@Birdy ~ %bash --version
GNU bash, version 4.1.5(1)-release (x86_64-pc-linux-gnu)

Basile Starynkevitch wrote:
> Hello All
>

> I almost released MELT plugin 0.9.6 for GCC 4.6& 4.7 but thanks to

[Basile Starynkevitch]

On Tue, Jul 31, 2012 at 06:52:17PM +0200, Pierre Vittet wrote:
> Hello,
>
> I have quickly compiled melt following your instructions (no problem
> for compiling/installing MELT).
>
>
> I tried to make and run ex01 and I guess I could reproduce the bug
> (or at least it produced one). I obtained a "The bug is not
> reproducible, so it is likely a hardware or OS problem". See the
> log.

Yes, that is the bug I'm tracking. GCC message is misleading, any plugin related bug gives the same
"The bug is not reproducible" message.

The bug is when running ex01 and the funny thing is that if you re-run the command

> gcc -fplugin=melt -fplugin-arg-melt-mode=nop \
> -fplugin-arg-melt-workdir=meltworkdir \
> -fplugin-arg-melt-extra=ex01m-helloworld -c empty-file-for-melt.c

which is done by make inside ex01 that command works well enough...

Perhaps some badly set variable in melt-examples/common.mk?

A big thanks for taking time to test that. That bug is making me scream...


Thanks, cheers

PS. I'm taking a 3 week break at end of this week.

[Basile Starynkevitch]

On Tue, Jul 31, 2012 at 04:30:00PM +0200, Basile Starynkevitch wrote:
>
>
> Hello All
>
> I almost released MELT plugin 0.9.6 for GCC 4.6 & 4.7 but thanks to Emmanuel Haucourt I found a bug
> which I cannot easily reproduce. Perhaps it is related to the differences (versions of make and of gcc)
> between our machines (Ubuntu vs Debian/Sid)
>
> In case you want to help and test it, you could get that thing from
>
> http://gcc-melt.org/badmelt-r190008-0.9.6-plugin-for-gcc-4.6-or-4.7.tar.gz
>
> the md5sum of that thing of 5606772 bytes is e636e1f7c4bb00b5f6c4a9d5b046167d
> it is extracted from MELT branch svn 190008
>
> (that URL is obviously temporary; I'll remove the thing when debugged the issue)
>
> What fails on one machine but works on another is the following human procedure
>
> Retrieve that tarball above
>
> untar it (it will produce a melt-0.9.6-plugin-for-gcc-4.6-or-4.7 directory without any badmelt in it)
>
> build the melt plugin as usual:
>
> make
> make install DESTDIR=/tmp/meltinst
>
> Then install it e.g. with
> sudo cp -v -r -p /tmp/meltinst/usr/. /usr/.
>
> Then get the melt-examples with git, e.g. with
> git clone https://github.com/bstarynk/melt-examples.git
> Then (a hello world example)
> cd melt-examples/ex01
> make all
> make run

I might have found the bug; when MELT is automagically building a module from C code (by running a make) the target module name was wrong. I might release 0.9.6.b (extracted from MELT branch svn 190018) in a few hours if my tests pass ok...

Thanks to all.