On 29.04.2011 01:45, Guenter Knauf wrote:
Hi,
Am 28.04.2011 11:24, schrieb Rainer Jung:
I hope I have some time to check older versions, like 1.3.9 etc. and
maybe also older apr (pool) versions to see, whether I can narrow down
the reason. Unfortunately until now, I could only reproduce the two
problems (unterminated loop, crash) when doing the testing as part of
the mass building, which takes time (a couple of hours). When running
testall after building even in loops, I could not reproduce the problems
...
I can only add that I get also sporadic segfaults (invalid mem access)
on NetWare with CodeWarrior compiler; and since otherwise httpd works
fine when comsuming same APR/APU I believe that something is wrong with
our test suite - it seems to me that it doesnt behave correctly in case
of errors, I mean that not allways return codes of a fucntion failure
are tested, and the test then runs into next function call with
paramters being NULL or something like that ...
also I believe that we fixed some of these already in 2.0 but not yet
backported;
though I had not the time yet to check into this deeper ...
anyway, I see no regression for NetWare - the failing tests are same as
with last release, so:
I did some more tests. There are various types of issues with
testreslist (unterminated loops, segmentation fault, bus error,
assertion failure). I get at least some of those even when using
- apr trunk as of now, so even with the volatile fix
- apr 1.3.9, 1.3.5, 1.3.2 (which was the oldest release I tried)
- apr-util 1.3.11, 1.3.10, 1.3.8, 1.3.2 (which was the oldest release I
tried)
using several combinations of apr and apr-util versions.
So if there is a common reason for those failures, then it is definitely
not a regression (and will also not be fixed by 1.4.3 as of now).
I will see whether I can strip down the test scenario and increase
reproducability in order to analyze. Since it is an old problem, it
should not block the release of apr-util 1.3.11 and apr 1.4.3.
Regards,
Rainer
