IRIX now requires the following defines: BROKEN_GETADDRINFO SETEUID_BREAKS_SETUID BROKEN_SETREUID BROKEN_SETREGID The BROKEN_GETADDRINFO definition should probably be tested for as it has been reported on the mailing list that 6.5.20+ might not need it. However, if it is defined, the openbsd-compat gai_strerror() function is used and its prototype conflicts with the definition on IRIX (const char * vs char *). Patch attached to fix this.
Created attachment 414 [details] Fix sshd on 3.7.1p1, gai_strerror prototype mismatch
AFAICT BROKEN_GETADDRINFO is not needed on IRIX 6.5.17 either.
BTW, for anyone else like me who is/was wondering if any of the other (than BROKEN_GETADDRINFO) three options can be left out, no, they all seem to be required as far as I can tell.
*** Bug 677 has been marked as a duplicate of this bug. ***
Patch #414 looks pretty good, but I have one question: is BROKEN_GETADDRINFO appropriate for all IRIXes? It'll disable IPv6 support.
According to the mailing list, BROKEN_GETADDRINFO isn't required on all IRIX systems. How about we leave it out for now. I'll try to come up with an automated test for a broken getaddrinfo.
*** Bug 691 has been marked as a duplicate of this bug. ***
Patch applied to HEAD and the 3.7 branch, with the exception of the BROKEN_GETADDRINFO part. Please update this bug when you have a working test. Can someone confirm whether or not those defines are required for IRIX 5? It's not included in the patch.
I just wanted to note that patch 414 does not address the issue that was reported in bugs 677 and 691. Given the situation that the 3.7 releases correct a security issue, it would be nice if us folks running IRIX were able to upgrade from 3.6.1p2. 677 and 691 are arguably duplicates of each other, but, in my humble opinion, neither are duplicates of 659.
...or at least it doesn't address the issue, until the realisation dawns that a patched configure.ac has no effect whatsoever on re-running ./configure && make, unless the changes introduced by the patch to configure.ac have been applied by some means to 'configure' itself. This was not obvious, but has now been filed away as an insight into the autoconfig stuff. Anyway, please disregard my earlier statement. :) Everything is now working as it should.
*** Bug 699 has been marked as a duplicate of this bug. ***
No good news for the daily snapshot (25/09/2003). There's still the problem on Irix (6.5.18f). We can connect on root account, but when it's another uid, it stop after the password and remains unavailable.
Please attach a server-side debug (eg sshd -ddd -p 2022) for your IRIX 6.5.18f box.
Created attachment 466 [details] Debug messages (Irix 6.5.21) Server-side debug messages for Irix-6.5.21 running openssh-SNAP-20030924.tar.gz
Created attachment 467 [details] Debug messages (Irix 6.5.21) Client-side debug messages for Irix-6.5.21 running openssh-SNAP-20030924.tar.gz
Luc, do you see the same problem without X forwarding?
no reply in 6 months == closed bug Reopen this if you want to restart the process with 3.8.1p1 when it is released.
Closed with release of portable OpenSSH 3.8.1p1