Work at SourceForge, help us to make it a better place! We have an immediate need for a Support Technician in our San Francisco or Denver office.

Close

#41 tuxcmd builds but fails to start on powerpc

closed-fixed
Tomas Bzatek
General (39)
5
2009-10-03
2009-02-04
No

Hi Thomas

I'm currently the maintainer of the tuxcmd package in Debian. tuxcmd has limited support only for little endian archs? Is that correct? Is also supporting big endian archs planed?

I recieved a bugreport that tuxcmd fails to start on powerpc (See: http://bugs.debian.org/513891\):

tuxcmd
malloc(93): 0x481FB010
free(0x481FB010)
malloc(93): 0x481FB010
free(0x481FB010)
malloc(93): 0x481FB010
free(0x481FB010)
malloc(93): 0x481FB010
free(0x481FB010)
An unhandled exception occurred at $BFB89110 :
EAccessViolation : Access violation
$BFB89110
$0FDE7E88
$0F9B6F04
$0FAD79CC
$0F8E9F90
$0F99B870
$0F9663E0
$0FDE0C5C
$0FDCCA88
$0FDCEC9C
$0FDEB980
$0FDEC9B4
$0FDECF6C
$0F9B70CC
$0FAE43B4
$0FDD2828
$0F9B6CD4

malloc(93): 0x48213010
free(0x48213010)
An unhandled exception occurred at $3D300A54 :
EAccessViolation : Access violation
$3D300A54
$1000CA6C
$10123640
$1011C50C
$1000DAB0
$10110EF4
$1000EF98
$1010FEE0
$1010FCAC
$1010FC18
$1010FBF8
$10110A94
$1011069C
$1000DD74
$1000BF54
$1000BDC4

I also attached strace logs of tuxcmd invocation. [I have no access to a powerpc machine]
If I can try to help or help provide more information, please let me know.

Thanks for your work on tuxcmd
Kind regards
Salvatore

Discussion

  •  
    Attachments
  • Tomas Bzatek
    Tomas Bzatek
    2009-02-04

    Hi Salvatore,

    yes, PowerPC port is currently broken. FreePascal has very bad support for different endianity, mostly from the memory management point of view. I had partial success some time ago on an old iBook G3 though, no plugins.

    FreePascal internally allocates memory from heap (most probably), but from different memory area than malloc does. It also adds few bytes control information and returns pointer with offset of few bytes, which obviously cannot be used with standard libc allocator calls. I had no luck with FPC cmem unit either which basically uses malloc, things went even worse.

    Spending so much time on fixing things which are broken from the beginning is unproductive and there are different, more important areas which I want to focus on first. Tux Commander could use complete rewrite to pure C, but that's fulltime work for several months. Volunteers are welcome of course.

     
  • Tomas Bzatek
    Tomas Bzatek
    2009-02-04

    Forgot to mention that I intentionally disabled PPC archs in the Fedora package. I can only support i386 and x86_64 at the moment.

     
  • Hi Thomas

    Thanks for your detailed reply. I think then the "only" thing I can go for at the moment is to disable build for the big endian builds. I'm looking forward to the development and will try to help where I can if possible.

    Kind regards

     
  • Tomas Bzatek
    Tomas Bzatek
    2009-10-03

    PPC issues should be fixed by commit 15efde21a039f2dfa11c715c9c40b6d12813baf3 incl. cleanup in commit 95dbf1260264603b86892530d6b581a35bdb4385. Please see commit logs for details.

    FPC > 2.2.4 is a must.

     
  • Tomas Bzatek
    Tomas Bzatek
    2009-10-03

    • labels: --> General
    • assigned_to: nobody --> tbzatek
    • status: open --> closed-fixed