IRIX and Software

6.5.30 troubles with inst and swmgr

Hello,

fresh install of 6.5.30.
swmgr crashes with a segmentation fault if the hist file grows past 4mb
inst goes a bit longer as in how big the file can be, but that's all

I've "fixed" it by moving the hist file to let it work on a new one but now even that doesn't help. I'm trying to install Mipspro , and after having opened all the cds and such, inst crashes with a segmentation fault with the smaller hist file present (600k) just after I write go and press enter.

If I remove the hist file, it goes as shown downward, then crashes due to segmentation fault.

Inst> go
Pre-installation check .. 3% Abort (core dumped)

Any idea ? Am I missing maybe some patch that's causing this behavior ? This is hardly fun...
Disk full?
Project:
Temporarily lost at sea...
Plan:
World domination! Or something...
not really... 240gb free....

Does it matter if the OS was installed while on an O2 R12k with 1gb ram, and then the disk was moved to a R5k with 256mb ? (should mention that initially sysmgr was crashing with a memory fault, and started working only after a requickstart all).

"heavy programs" (jaleo/effects, by example) work without issues
A bit more data, but is not helping me a lot... maybe gives an idea to someone else

Code: Select all

sgio2 15# dbx /usr/sbin/inst core
dbx version 7.2.1 Mar 13 1998 19:00:36
Core from signal SIGSEGV: Segmentation violation
(dbx) t
>  0 <Unknown>() [< unknown >, 0x117bb3a4]
(dbx) wi
[???, 0x1185d508]      addiu   s1,a5,2
[???, 0x1185d50c]      ld      a2,0(sp)
[???, 0x1185d510]      lw      a7,0(s0)
[???, 0x1185d514]      addu    a7,a7,a4
[???, 0x1185d518]      lw      a7,0(a7)
*[???, 0x1185d51c]      sw      a7,0(a2)
[???, 0x1185d520]      lw      a6,4(s0)
[???, 0x1185d524]      addiu   a3,a3,1
[???, 0x1185d528]      slt     a6,a3,a6
[???, 0x1185d52c]      addiu   a4,a4,4
(dbx)
a "where" might give some more usable info
r-a-c.de
where returns > 0 <Unknown>() [< unknown >, 0x117bb3a4]
okay no luck there.

well, how is the system otherwise? is that the only problem you have?
anyway since this is a fresh installation the fastest method would be installing it again. on the target machine this time of course.
r-a-c.de
Aside that, the machine is stable.

Anyway I "think" I managed to fix it, at least up to now it didn't crash again. The only annoying thing is that each rqsall at the end of inst/swmgr sessions, now, takes 100 to 109 minutes with the cpu past 90%.
Is there any trick to make the whole rsq/rsqall snappier ?
how did you fix it?
r-a-c.de
well, I found out that there was a mipspro 4.2.2 installed , and 4.4 partial runtimes, and so on.
First I've used the overlay to mipspro 4.4.4m, then I've kept re running all the sgi made cds I had until they all were showing "same version" or "older version" and 0 conflicts.
Adding the subdirectories (dev) of the 6.5.30 apps / complimentary apps helped bigtime (to the point that all Cosmo Suite parts installed with 0 conflicts with the most recent java)
Pity for the o2 demo cd (damaged), and the lack of the mekton demo
mazzinia wrote: well, I found out that there was a mipspro 4.2.2 installed , and 4.4 partial runtimes, and so on.
First I've used the overlay to mipspro 4.4.4m, then I've kept re running all the sgi made cds I had until they all were showing "same version" or "older version" and 0 conflicts.

you mean 7.x. there never was a mipspro 4.4.x.
anyway what do you mean by "found out"? i thought the installation was new?
or did somebody else install it?

Adding the subdirectories (dev) of the 6.5.30 apps / complimentary apps helped bigtime (to the point that all Cosmo Suite parts installed with 0 conflicts with the most recent java)
Pity for the o2 demo cd (damaged), and the lack of the mekton demo

seems there's a lot installed. if there's a problem, it's usually a good idea to slim down as much as possible.
anyhow i don't see how installing some dev stuff would fix a crashing inst.
r-a-c.de
err yes, 7.x ... mispelled. The installation was new but not done by me. I think it was straight 6.5.28 and then overlayed to .30. Could be that something went wrong while overlaying.

And don't ask me how the dev fixed it. The fact is that it has solved all the problems, anyway. I really only would like that the rsqall would not last that long ( suppose that with a 300mhz r5000, the time would go down by 50% due also to the cache )
mazzinia wrote: I think it was straight 6.5.28 and then overlayed to .30.

no, there is no "straight". everything but 6.5.0 is an overlay but you can install an overlay in one go together with the 6.5.0 base release.

I really only would like that the rsqall would not last that long ( suppose that with a 300mhz r5000, the time would go down by 50% due also to the cache )

yeah this is way too long. given the quite big inst index and the problems you had (and still have) i'd assume that the installation is messed up somehow.
as mentioned before, i'd re-install it and go for a rather light selection at first. then, later when irix is running you can install more
r-a-c.de
mazzinia wrote: Is there any trick to make the whole rsq/rsqall snappier ?

If you're not already past the point of no return, there's some r qs -related advice from davea (the person who maintained rqs/rqsall for SGI) in this post: viewtopic.php?f=10&t=10959&start=15#p85984
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
define the point of no return ? :)
Thanks for the link, but I think I read it a bit ago, when I was looking to fix the initial issue ( a manual full rqsall as in that thread fixed it ).
Now that I think about it, a full rqsall took over 3 hours (but I may be wrong)

On a side note, the "fix" created a side effect : inst / swmgr -X/ shows NA for installed programs that are in hist, now (while launching without the -X/ allows to see dimensions etc, for such programs).
Weird, I would say. inst/swmgr 4.1 is the latest ? no patches ?
Is there any patch that's required to prevent issues, on 6.5.30 ? (in case one such thing is missing)