I'm building an image on i386 (little endian) and running it on big endian arm. This is failing though. Creating the image: [ 5:42PM] hamish@snake:oboe/utils/fwmgr $ dash /usr/bin/makeself img bl.run bootloader installboot Header is 375 lines long WARNING: Overwriting existing file: bl.run About to compress 4144 KB of data... Adding files to archive named "bl.run"... ./ ./ramdisk.gz ./zImage ./installboot CRC: 1373132883 MD5: 9e2c27a8a140411b558566d636f86cf4 Self-extractible archive "bl.run" successfully created. [ 5:42PM] hamish@snake:oboe/utils/fwmgr $ md5sum bl.run be6dc69e124bbd9307b10301a7962b51 bl.run [ 5:42PM] hamish@snake:oboe/utils/fwmgr $ sh bl.run Verifying archive integrity... All good. Uncompressing bootloader.... bl.run: line 365: installboot: command not found (Extracted ok although the script was missing; not to worry.) Now copy it to my target and run it; # md5sum bl.run be6dc69e124bbd9307b10301a7962b51 bl.run # sh bl.run Verifying archive integrity...Error in MD5 checksums: d41d8cd98f00b204e9800998ecf8427e is different from 9e2c27a8a140411b558566d636f86cf4 Hamish
This doesn't seem to be a cross architecture thing after all. A .run created on amd64/unstable runs ok on ppc/sarge (brucker.debian.org). My target is running busybox for its shell, coreutils etc which may be the cause. I'll keep investigating, Hamish
Hello. I tried to reproduce that bug on SPARC which is AFAIK big endian and had no luck. Does it still exist in 2.1.5 version? regards fEnIo