Werner’s own blurbs
CCID reader fixes
March 14th, 2005
Due to changes in Linux 2.6, gpg’s internal CCID driver did not worked anymore reliable. Finally I took the time to browse over the USB stack code and to learn the basics of USB. Despite several hours of debugging, I was still not able to locate the bug there. Apparently it is a problem with the SCM readers depending on the way they are accessed. It seems to be pure luck that it worked on 2.4. The fixes I implemented are to check for card presence before powering up the card, progressive timeout retries and resetting the device right before releasing the interface. So far it works for me. Its in the CVS and I try to add a patch file my home directory over here.
1504 is a bad size for some keys
March 13th, 2005
It turned out that using keys of certain lengths which are not a multiple of the 8 bytes may cause a segv only under Windows. Timo tracked the problem down to the generation of the session key and I am now looking for the real cause …
fixed a long standing bug
March 13th, 2005
After quite some hours of debugging using plain old printfs and staring at the output, I found a long standing bug in the secmem_realloc function which exhibit itself only with certain request sizes due to padding and blockreuse issues. This fixes the 1504 bug I mentioned yesterday (well due to problems with the blogging code the date of that entry is possible wrong).