Pekka Laukkanen wrote: > Some time ago I asked about the limitation that one person (actually a > process) can be member of only 32 groups. The problem was originally in > the kernel but ought to be solved with the 2.6 release. The problem > still exists in Debian Sarge with 2.6.8 kernel, though, as my original > message (included below) shows. The problem seems to be rather hard > since nobody replied. Just today we needed to do some extra work because > of it and I decided to ask again. I saw your original message and noted it here as a warning that I might run into it in the future. It was useful to me just for that purpose if for nothing else. > This time I would also be happy if someone could point me to another > mailing list (or other source) where to ask. Would it be appropriate to > use for example debian-devel? As an extreme measure I would also be > interested to hear about other distros that would not have this > limitation anymore. If you are motivated I would suggest getting the pristine kernel.org linux-2.6 sources and compiling a stock kernel without the Debian patches. Then see if the behavior persists. If it works then you know it is something Debian has done with patches. If not then it is an upstream problem regardless of the changelog entry that says it was fixed. This exercise needs to be done to isolate where the issue lies. At that time I would take the discussion to the linux kernel mailing list and ask there. It might actually not be fixed or it might need some additional special configuration that is not being done. One thing to be careful of is that utilities such as 'groups' may have limitations in the code. I don't know. But it would certainly be possible for a utility to have a smaller buffer or expect a different behavior than the kernel. If you see something that looks like a fixed buffer size in one of the coreutils please be sure to report that to the the bug-coreutils mailing list because it is not intentional. Bob
Attachment:
signature.asc
Description: Digital signature