erg@ulysses.UUCP (07/09/87)
Release 9.5.1 has more than just tcp/ip problems. The new
version of Domain/IX broke things that have worked for years. Amidst the general
damage, some bugs have made themselves obvious:
1) In BDS4.2 mode, if you set a handler for the SIGCHLD fault, you get
the System V SIGCLD signal instead.
2) The new system version of malloc assumes it is the only one going to call
sbrk, causing dire problems if your program tries to do some memory management
on its own.
3) Berkeley soft links no longer work, noted by Apollo under "Known bugs".
4) Pipes can have holes; it's possible to lose data through them.
We also now have mouse problems. We use our own mice,
attached to an RS232 port, with motion and button hits
reported to the operating system using the appropriate smd calls. This gives
us much greater flexibility and power for constructing user interfaces than
Apollo's mouse. In particular, various services such as (efficient) pop-up
menus are available at the Display Manager (DM) level, and can also be accessed by
otherwise non-graphical programs. In the new release,
although motion events are received by everyone, and mouse button hits are
received by the DM, application programs don't necessarily
receive the hits, breaking lots of programs.
Ironically, we moved to 9.5.1 because neither we nor Apollo could
get our ethernet working under 9.2. It turns out that under the old software and
hardware, e.g., tcp 2.1/SR9.2 and the com-eth board, introducing
802.3 ethernet equipment onto your ethernet can cause what Apollo
calls the "bogus bit" problem, rendering your Apollo useless on
the network. To solve this, Apollo recommends upgrading to the ecmb
ethernet board and tcp 3.0, which also requires SR9.5. Well, tcp 3.1's
routing software is quite broken - running routed on our server causes
all the network software to break, apparently due to the large volume
of routing traffic. Apollo's "fix" is to use static routing.
We need to reach distant machines that we have
little administrative control over, so Apollo's
solution of static routing is of little consolation to us. Essentially,
all of our Apollos have been isolated from the rest of the world
for over two months.
We have reported all of these problems and more, but no one has offered us
an early release of 9.6.