Mozart FTBFS (#332630)

[Kevin Glynn]

Hello Porters,


My package mozart FTBFS on s390 (see Bug#332630). The Mozart package
first builds an emulator (written in C++) and then uses the emulator
to compile the Oz compiler. The first file we try to compile is
giving a dynamic type error. Since this works successfully on
(nearly) all other platforms I am thinking the emulator must be
corrupt.

I am seeing similar issues on m68k, which I have access to but
progress is slow (e.g. package builds fine with gcc 3.3 but fails with
most recent gcc 4.x).

Is it possible to get access to an s390 machine. I'm not a DD but
have a DD signed key.

Thanks for your help,

Kevin

[Bastian Blank]

Post by Kevin Glynn
The first file we try to compile is
giving a dynamic type error. Since this works successfully on
(nearly) all other platforms I am thinking the emulator must be
corrupt.

It is not built on powerpc, even if it says, it should work.

Possible problems:
- char is unsigned.
- the build shows implicit declerations of the builtin functions strlen,
strcpy, memcpy.
- "warning: comparison is always true due to limited range of data type"

Post by Kevin Glynn
I am seeing similar issues on m68k, which I have access to but
progress is slow (e.g. package builds fine with gcc 3.3 but fails with
most recent gcc 4.x).

Ah, that looks like problems with undefined behaviour.

Post by Kevin Glynn
Is it possible to get access to an s390 machine. I'm not a DD but
have a DD signed key.

I'm afraid, but the s390 machines are developer only.

Bastian

[Bastian Blank]

Post by Bastian Blank

Post by Kevin Glynn
I am seeing similar issues on m68k, which I have access to but
progress is slow (e.g. package builds fine with gcc 3.3 but fails with
most recent gcc 4.x).

Ah, that looks like problems with undefined behaviour.

It is something like that. It works with -O0, hangs with -O1 and
completely fails with -O2 and -O3. Also the shared libs needs to be
built with -fPIC instead -fpic.

Bastian