Joachim Schrod <jschrod(a)acm.org> writes:
================================================================
Dear Bug Team!
When one updates the package index via #'package-get-update-base, with
verification by GPG, approx. 1 out of 5-10 times the system tells `The
message is corrupt'.
I use gpg 1.2.5.
Frankly, I don't believe that the fetched index was really corrupt. I
append an EFS buffer where it happened at the 5th try. The amount of
bytes that got transferred is the same as in the previous times where
verification succeeded. When one tries again, it usually succeeds
again -- once, I had two `corrupt message' in a row.
I haven't found a method to tell EFS that it shall keep its temporary
files in /tmp/$USER; it would ease the analysis of this problem. Is
there a variable somewhere? aproposing efs didn't show anything
obvious.
Cheers,
Joachim
I suppose this could possibly be caused by the bug just fixed by David
Evers: flush the last chunk of process I/O, even when the pipe is
blocked.
You might want to try your test case against the latest 21.4 CVS
sources.
HTH,
Vin