A Call To Stop Using Uclx Compressor

[iamasmith]

OK, below were my initial thoughts on UCLX compressor after I thought that I had effectively sized the /dev/shm device to produce a /tmp capable of 10Mb storage.

I since found out that the /root/etc/rc.d/rc.rofilesys carries the mount command for /dev/sdm in about 3 locations.

Thanks for calling my attention back to this chrget.

The problem IS indeed related to the size of /tmp



Original posting.

UCLX compressor stub crashes on PXA270 models with a bus error. Simple as that.

If anyone is considering packaging anything at all into IPK then please consider portability to the PXA 270 models.

UCLX compressor is written in ARM Assembler and I'm not sure the project is being maintained any more. inter.zon@free.fr is a dead email address.

Until we can get this nailed then please take heed.

If you do have a package that is throwing bus errors then it is worth attempting to decompress the binaries with uclx -d. If they expand then there's a good chance that will fix it.

Known effected packages : Kismet (kismet_client and kismet_server are both compressed with UCLX).

- Andy

[chrget]

UCLX compressor stub crashes on PXA270 models with a bus error. Simple as that.

Hmmm, are you sure it's a problem with the CPU? The UCLX decompressor stub will indeed produce bus errors -- but not just on the PXA270. Even on the good old SA1110 this will happen, if there is not enough space (i.e. enough space for the uncompressed executable) in the filesystem carrying the /tmp directory.

[...] Until we can get this nailed then please take heed.

Indeed, packaging UCLX-packed executables is a bad idea. It should be left to the user if he wants to pack his program files or not.

Best regards,
Chris.

[iamasmith]

chrget, thanks, you are indeed correct. If you look at the modified first posting you will see what the problem was.  

Regards,

Andy