It's raining again.
Spent the day sorting out where I'll be working for the next six months. Not entirely happy, but work is work.
Got home to find a parcel had arrived. Thought it might finally be my copy of Solaris ... nah, 'fraid not.
Made a dent in the email backlog from last night, and tried to catch up on some missed sleep.
I'm in the ten-pin bowling team (by default, I hasten to add! It certainly isn't on ability) at work tonight, so I won't be around to work on GNQS.
Bit of a mixed day today ... I don't know where (or if) I will be working come Thursday morning ... won the Doubles Final in the bowling league ... can't connect to my ISP to get at my email (or upload the diary ;-) ... still can't find the fault which is corrupting data on the local LAN :(
Jack Perdue has kindly looked at v3.51.0-pre6, and highlighted a problem with the way SETUP works out what platform it is running on. This one is quite serious, as it prevents -pre6 compiling out of the box on HP-UX v11.
We had another power cut last night ... looks like it took out the whole estate ... about 40% of the way through uploading Generic NQS v3.51.0-pre6 ;(
Anyway, it's not raining today, so I'm off for a lovely hike over some lovely Welsh hills with some lovely Welsh company. I'm not sure it's going to be a good day for taking photos, but what the heck ;-)
-pre6 has been uploaded (and announced), so I hope you download it and give it a try.
7:30am is far too early to be trying to fix anonymous ftp to ftp.gnqs.org. Turns out the problem was related to the shared libraries in the chroot'd environment ... why isn't ftp's ls command statically linked?
Anyway, -pre5 is definitely out. -pre6 will follow later today. The only difference between the two is that -pre5 will not include the prologue/epilogue fixes, -pre6 will. I'm not confident that the prologue/epilogue fixes will work at all first time (although I have tested them here first ...) so -pre5 is available just in case.
Here's how the new prologue/epilogue stuff now works. This is a refinement on Thursday's approach after working with the code some more ...
This approach sees a big simplification of the way we start batch jobs, at the expense of a couple of extra shells. Oh, and we get a lot of extra power as a result too.
No GNQS work today (apart from answering a few emails) as I had the insurance man round helping me fill in lots of nasty forms.
Hrm ... I don't know what happened to Wednesday ... I spent most of it asleep :(
Anyway, Thursday ... making good progress in fixing all of the compiler warnings with -pre5. Now I can get back to making -pre5 releasable.
Jack Perdue at TAMU has been trying -pre4 on various machines, and found an important problem with SETUP. If you run SETUP on HP-UX 10, the best it can do for the defaults is load the platform info for HP-UX 9. This is not a good idea, given the differences in how GNQS compiles on the two platforms.
Question is - how to deal with this? I'll fix it in -pre6 ... it's more important to get -pre5 out for testing this time, and then make -pre6 ReleaseCandidate #1.
And confirmation from Jack that GNQS v3.50.4 + the HP-UX patch (which is now in -pre5) appears to compile and work on HP-UX 11 ;-) I'll have to check to be sure, but it would be nice if we were first with support for this new platform.
Prologue/epilogue support ... I've decided how to do this.
GNQS calls prologue script, which runs as root.
Prologue script does what you code it to do.
Prologue script executes nqsexecjob.nqsexecjob changes uid to the uid of the user who owns the job.
nqsexecjob then runs a second prologue script.
nqsexecjob then executes the actual job.
nqsexecjob then runs the epilogue script.The prologue script now doubles as an epilogue script, allowing the sysadmin to do whatever they believe they need to do.
I'll supply a default root prologue script (we need a better name for it ...) which you can customise. If the inner prologue/epilogue scripts are not present, then nqsexecjob will simply not try to run them, and just run the user's actual job.
It should be simple to code (famous last words ;-) and also provide the flexibility we all want on this. All of the scripts have the right environment, and the user's job and inner scripts can inherit environmental changes from the outer script - an important feature.
More news on the holiday front ... I'll know by Wednesday next week where I'll be working for the next six months, and therefore when my long-awaited holiday in October will be. I have every intention of being uncontactable for a fortnight (no change there I guess ;-) but I'll make sure everyone knows before I do.
I'm having one of those days where the best solution is to just go back to bed, get up and start the day again.
It seems that the type checking on egcs, the C compiler shipped with RedHat Linux 5.1 is somewhat stricter than the older GNU compilers. I'm not arguing that it's correct - I guess today I'd just like to be able to get GNQS compiled and released without this.
Sigh.
It also appears that those individuals who are currently trying to break into www.gnqs.org again haven't given up. If they've actually broken in this time, I can't tell, but I don't think so. Doubtless they will be back.
I doubt they're reading this, but just in case ... I'd like them to bugger off and leave this machine (www.gnqs.org) alone. Their activities are taking time away from work on free software. It's one thing to break into a box - it's another to use it as a staging post to attack other boxes.
On the positive front, (and I hope she doesn't mind me saying this ;-) it looks like GNQS will soon have a volunteer working on the documentation. It's not clear yet whether she'll be working on it full-time (funded out of my pocket) or donating her time as I do, but either way I sure am looking forward to when she gets started. She's a professional writer ... I've known her for the last six years now, and she's a good writer too.
The current timescale is for her to start working on GNQS in January. She's not familiar with GNQS, so don't expect overnight results, but I hope we'll all find this a positive move, and a good step forward.
Mmm ... both films last night were wonderful. The Horse Whisperer just drips with power and emotion backed up by great photography and a great cast, and in Cube, it's great to see a good sci-fi horror flick get an airing before disappearing without a trace. It's worth seeing for the opening sequence alone ;-)
No work today on Generic NQS ... spent the day putting www.gnqs.org back together. It looks like someone has been trying to break into the box once again, but so far there's no evidence they actually got in this time.
Anyway, email to stuart@gnqs.org should now be working again, and hopefully by the time you read this, www.gnqs.org will be working again too.
I'm off to Cardiff this afternoon, to see The Horse Whisperer and then the special showing of Cube at the Odeon which starts very late. I don't expect to get much done on GNQS tomorrow as a result.
No CD of the week this week - a video recommendation instead. Absolute Power, starring Clint Eastwood, Gene Hackman, and Ed Harris, has finally come out to buy here in the UK. It's a strong story of a burgler who witnesses a murder involving the President of the USA, and how it plays out. Mmm ;-)
Still working on the prologue/epilogue stuff, in between playing with Enlightenment [more] and KDE [more]. Unfortunately, the Internet seems to be a little unhappy this morning, resulting in plenty of unreachable hosts and dropped connections. Sigh.
I'm also integrating in a memory checking library called MSS [more], originally developed for DJGPP [more]. I've made a couple of changes to make it compile as part of GNQS, and also to ensure it doesn't leave a file descriptor open between log messages.
The memory checking will be a compile-time feature. I don't recommend anyone tries this - it could prevent GNQS working properly until I've ironed out any integration issues - but it's there to try if you keep getting serious crashes of GNQS.
SETUP now supports a number of new command-line switches too, which will turn on SETUP debugging support, and which can be used to start SETUP at specified stages of its work.
Some kind soul has broken into www.gnqs.org, and tried to use it for purposes it is simply not there for. This has resulted in the machine being isolated from the network whilst we rebuild it.
Which means I'm talking to myself for the next few days until the box is back online ;-)
It also means that GNQS v3.51.0-pre5 will be delayed until I have an FTP site to upload it to. To be honest, this is not a bad thing; I didn't go back to work until yesterday, and so haven't been able to even look at GNQS all week :(
The prologue/epilogue stuff still isn't finished, as I'm trying to make it so you can have scripts which run both as root, and a set which runs as the user who owns the job. As soon as I get something working, I'll make the pre-release available, and we'll iron out the rough edges for the next pre-release.
Off work with the 'flu' :(
And www.gnqs.org is temporarily unavailable while we investigate a suspected hacker attack. Which isn't going to happen for a couple of days until I'm back on my feet.
CD of the week ... I bought Ghostland's self-titled debut several months ago now. It doesn't really have a single outstanding track, but it works very well as a complete work, and has managed to claim a semi-permanent slot in my hi-fi as a result.
Work continues on GNQS v3.51.0-pre5.
It now compiles once again on my laptop. I'm now trying to incorporate the following:
AFAIK, HP-UX v10 is still the most broken platform GNQS attempts to compile on. I've been sitting on fixes kindly contributed by the GNQS user community for quite literately a year now; I'd like to get them integrated before releasing -pre5.
The current implementation of prologue/epilogue scripts in GNQS isn't very good. Ideally, we need to run the prologue script, then the user's actual job, and then the epilogue script, all from the same parent shell.
I've had a look at how the folks at CERN do this; I like the principle, but I'm not sure about the way it chooses which shell to use for the prologue and epilogue scripts. I think I'll continue to force both scripts to run as /bin/sh scripts.
Daniel Israel suggested a couple of changes to SETUP to reduce the number of recompiles it causes, whilst ensuring that binaries for the wrong architecture cannot be re-used easily.
These changes are also in -pre5.
I've still not installed FreeBSD, but Tony Maher has contributed a number of fixes and suggestions which need incorporating.
SETUP also needs a few more tests for various little things like SVr4-specific libraries; this may wait until -pre6 in a fortnight's time.
A week of frustration continues.
The ISDN service to my ISP at home has been ... intermittent at best.
I've spent so much time answering email that I've not finished making -pre5. (That's been in a good cause tho - more in a mo)
My attempts to get data off my desktop PC without corruption have failed miserably. To and from the local harddisks is fine. From CD to harddisk is fine. Harddisk to CD is not fine. Harddisk to network is even worse. Harddisk to parallel port doesn't work either.
The latest books and CDs I have on order haven't even been dispatched yet, and there's no sign yet of Solaris/Intel arriving.
Oh, and I didn't get the job.
However, it's not been all bad.
Negotiations with the technical writer continue. It's possible the writer won't be available to work full-time on the manuals after all, but if not, they are interested in contributing their spare time to such a project anyway. No guarentees until we have a manual to publish, naturally, but at least things are moving in the right direction.
I've also been having a fascinating chat with a long-standing GNQS user about the possibility of some modifications to allow scheduling of parallel tasks. This would be a great way to finish off the v3.5x source tree ;-)
I've started up a Developer's Manual for GNQS v4 here. This will be where we publish our requirements, and then the evolving high-level design. The plan is to publish the latest version of the requirements and design to NQS-Developers fortnightly, to keep the discussion going.
(Yes, this does mean that there will not be a Developer's Manual for the current GNQS source tree.)
GNQS v3.51.0-pre5 still doesn't compile :( The job interview could have gone better as well.
Only a half-day on GNQS today, as I'm off over to Cardiff this afternoon.
Still working through the problems w/ GNQS compiling on RedHat 5.1. SETUP isn't passing the default switches to egcs from the platform-config file either ...
Spent quite a bit of time working on WASP (yes, it now has a name ;-), the web-based helpdesk software I'm now working on for Generic NQS. We've finished the list of features we need, and the basic database has been designed. Work has started on the code, but don't expect anything visible for quite a while.
I have a job interview tomorrow ;-) It also means I've got tomorrow off work - I can spend the rest of the time on GNQS.
The car is going to cost several hundred squids to fix :(
More work looking into the cause of some problems on HP-UX v10.20. The error message which comes out should only appear under one circumstance - if the parent process ID is either 1 or 0 (i.e. netdaemon's nqsdaemon is missing). I've changed the code here to provide an explicit error message ... we should know soon whether this is the fault or not.
To be honest, it's about time I sorted out the HP-UX port. It doesn't compile out-of-the-box.
-pre5 doesn't compile since the upgrade. There's a problem with sys_errlist[] in the lpserver which I'd missed when fixing all the other occurances. Looking at the SETUP scripts, tho, I'm not sure that I'm testing for this correctly in the first place ... I've fixed both lpserver.c and SETUP; we'll just have to see whether I've broken any ports or not.
Hrm. Looks like lpserver isn't the only problem. Should be fixed in a day or two.
Nothing achieved on GNQS tonight ... mainly because the nice people from the AA had to tow me and my car home :(
Somehow, I ended up with my wiper blades tangled together ... this burned out the motor. This is a bit of a problem when travelling at night through our lovely English rain.
Looking into a problem with HP-UX v10.20; the netdaemon doesn't run, and to make matters worse there's no output to syslog at all (not even the GNQS startup messages!).
The syslog will hopefully be a problem with the syslog.conf file, or (more likely) the daemon hasn't been restarted. Netdaemon ... that's going to take a little more digging.
My thanks to everyone who has sent in testimonials so far to go on the new web site. Your support, as always, makes all the difference.
Well, that's www.gnqs.org finally uploaded and announced.