loadup-apps-from-full.sh fails on WSL-Pengwin but not WSL-Ubuntu

[fghalasz]

loadup-apps-from-full.sh fails on WSL-Pengwin but not WSL-Ubuntu. Failure is a break in NC.InitializeUID.

[fghalasz]

@masinter @pamoroso
Just checked in a fix to Notecards (Notecards PR#14) that should fix this. Issue was that in some cases \MY.NSADDRESS was NIL, causing NC.InitializeUID to break with an illegal arg. Fixed NC.InitializeUID so that if \MY.NSADDRESS is NIL, it uses RAND number instead for setting the UID base.

Please close this issue if this fix works.

[nbriggs]

I tracked down this code in Maiko --

/************************************************************************/
/*              i n i t _ i f p a g e _ e t h e r                       */
/*                                                                      */
/*      sets Lisp's idea of \my.nsaddress. Clears it if ether not       */
/*      enabled                                                         */
/*                                                                      */
/************************************************************************/

void init_ifpage_ether(void) {
  InterfacePage->nshost0 = (DLword)((ether_host[0] << 8) + ether_host[1]);
  InterfacePage->nshost1 = (DLword)((ether_host[2] << 8) + ether_host[3]);
  InterfacePage->nshost2 = (DLword)((ether_host[4] << 8) + ether_host[5]);
}

would it be preferable to leave the Notecards code alone and fix init_ifpage_ether so that if the MAC address has not otherwise been set it creates an address in the "locally administered" space (02:xx:xx:xx:xx:xx) ?

[nbriggs]

Another alternative -- call (\SETLOCALNSNUMBERS) if you need to use \MY.NSHOSTNUMBER and it's NIL. That sets up a generic 0.12345.0 (octal) default address.

[pamoroso]

@fghalasz Which branch of the Notecards repo is the fix in? fgh_fix-NC.InitializeUID or main? Should I try building Medley from main or some other branch?

[pamoroso]

With the following branches and heads:

• Medley: master 5b37dd0
• NoteCards: main fa2323e

the build still fails with a stack overflow on Linux Mint 22.

[fghalasz]

@pamoroso Ok. I'll keep looking.

@nbriggs I think I'll keep the change to Notecards since it's the right thing to do in terms of system stability. But should also go ahead and make the maiko changes. I would prefer a solution where the NS address isn't the same for all machines since in Notecards it's used to create a globally unique id.

[MattHeffron]

Would it make sense to construct the default NS address from the MAC address of the first non-loopback interface? That's protocol-independent.

[nbriggs]

@MattHeffron well, yes -- because the NS address is the 48-bit MAC address of a system. If you compile with ether access enabled it will try to pick that up. Currently it's only enabled for SunOS/Solaris, where we have NIT and DLPI code for the network - I started on doing it with a more machine independent (at least in maiko code) way using pcap, whatever one does, though, one needs to either be root or a member of a group with access to the network interface to be able to pick up the address.

[nbriggs]

(or perhaps not root if it's only to read the address -- I'll check on that)

[masinter]

with what is checked in as master for maiko and medley repos and main from the notecards repo, i can do a loadup-all.sh --apps without any failures. So I thinkthe problem that started this has been fixed (by changing notecards so it doesn't fail if \MY.NSHOSTNUMBER is NIL).

Other problems should be discussed in the appropriate issue / PR.