[OpenAFS] problems with file links under Windows XP
Thu, 24 Jan 2008 15:44:22 +0100
I try to produce a minidump and a log file with Dr. Watson. A public=20
test case isn't possible, because we don't have a public cell.
I agree with you, that it make no sense to create such links. But our=20
software developer use scripts for creating links with absolute paths to=20
include and library directories. The results of some scripts are such=20
links with "../.." and "..". I also think, that when such links are=20
possible under Linux, it could also happen by other users.
The limit of 100 characters is really strange, but I did some test with=20
shorter links without problems.
When you have clients to test, I can do an install and check these=20
clients against our environment.
Jeffrey Altman schrieb:
> If you have a reproducible test case, please follow the directions in
> the release notes and file a bug report to firstname.lastname@example.org.
> In particular, if you can place a live public test case that I can run=20
> a client against it will make things much easier to track down.
> Also, I would really like to understand the rationale for creating
> a link that includes either ".." or "../..". Can you provide me an=20
> example where this makes sense?
> The limit of 100 characters seems very strange. The max length of a=20
> target is 1024 characters.
> Thank you.
> Jeffrey Altman
> Lars Richter wrote:
>> Hello all,
>> we get some serious problems with file links in our environment.
>> We are using OpenAFS for Windows version 1.5.30 and Linux AFS-Servers=20
>> When we create a file link under Linux with a pattern like=20
>> "PATH/../../PATH" and the path is longer than 100 characters, we get=20
>> an error on Windows "Permission denied". In a 64bit environment the=20
>> AFS-Client crashes with a message "OpenAFS Stopping due to error=20
>> (cm_scache.c:1783): cm_scache_t refCount 0." in the EventLog.
>> I did some tests with older versions. The result is, that the latest=20
>> version without the error is 1.5.21. I think, there is a problem in=20
>> the recursion detection feature in version 1.5.22, which was partly=20
>> fixed in 1.5.23 and 1.5.24.
>> Does the path have a pattern like "PATH/../PATH" there are no problems=
>> Does anyone have any suggestions?
Dipl.-Inf. Lars Richter
Pro Design Electronic GmbH
Europaplatz 5, 99091 Erfurt
phone: +49 361 55038-25
fax: +49 361 78930-80
Helmut Mahr, Ulrike Angersbach, Stephan R=F6slmair, Heiko Mauersberger
Registergericht: Amtsgericht Traunstein
Registernummer: HRB 13 002=20