search

PDP-10 / klh10

Community maintained version of Kenneth L. Harrenstien's PDP-10 emulator.
Other
59 stars 7 forks source link

CH11: fix long-standing data corruption bug #54

Closed bictorv closed 3 years ago

bictorv commented 3 years ago

Instead of accepting new input packets after checking that the input packet is shaped OK, do it after actually reading the packet to ITS. Otherwise read packets can be overwritten while only partially read to ITS, leading to corruption of data.

This happened much more often on Eric's virtualized systems than on my Raspi-3-based one, and much more rarely on UP which runs on an old Raspi B. But now there will be order in Chaos.

larsbrinkhoff commented 3 years ago

The FreeBSD CI build is failing. I had this problem with ITS too, and I found no easy fix.


./autogen.sh
autom4te-2.69: need GNU m4 1.4 or later: /usr/local/bin/gm4
bictorv commented 3 years ago

The FreeBSD CI build is failing. I had this problem with ITS too, and I found no easy fix.

./autogen.sh
autom4te-2.69: need GNU m4 1.4 or later: /usr/local/bin/gm4

Hmm. Install GNU m4 v 1.4 or later?

larsbrinkhoff commented 3 years ago

As far as I can see /usr/local/bin/gm4 is GNU m4 1.4.

bictorv commented 3 years ago

So how does autom4te check the m4 version? (I have no clue about how CI works here!)

larsbrinkhoff commented 3 years ago

I don't know. It could be that "gm4" fails to print some expected output, or even crashes.

Cirrus CI is used to build for FreeBSD. It uses the .cirrus.yml file for configuration. It first runs the commands in "install_script", and next "script". I would think the ./autogen.sh part fails.

Rhialto commented 3 years ago

Maybe gm4 was removed from the build environment as it is supplied by cirrus? (just guessing...) because this always worked, nothing changed on the klh10 side, and the Travis build still works.

Rhialto commented 3 years ago

We never discussed merging procedures I think :) Shall the author of the merge request merge it, or one of the approvers, or even I as the former owner of the repo? I know that in gitlab you can see who is allowed to do the merge, but I haven't found it in github yet, so it could even be that not so many people are allowed to do it.

larsbrinkhoff commented 3 years ago

I think the commonly agreed default is to let the maintainer merge when the pull request is in good shape.

For the ITS repo, I like submitters to merge when they have approval.

Rhialto commented 3 years ago

I'll merge it then. It's just that for example at $WORK, we have the procedure that the author merges (and then takes the work of rolling it out) , after the MR has been approved by at least one other person.

larsbrinkhoff commented 3 years ago

Are you using Gerrit at work perhaps?

The question only comes up because I like to invite people to the PDP-10 group and give out write permission to the repositories. In the more common case only repository owners or contributors can click the merge button. But there's no defined policy, so let's go with the "default" unless there's a reason otherwise.

Rhialto commented 3 years ago

No, we're not using gerrit at work (but I know it from openstack context). At work we use gitlab (on-premises).

I did find a page about access to the repository but it confused me. Either everybody in the PDP-10 organisation can merge/commit/etc, or just me. Or they listed both because they may see that as different independent groups of people.