On Sun May 13 1:49 , 'James G. Sack (jim)' <[EMAIL PROTECTED]> sent:
>Ralph wrote: >>.. >> 3195 ? S 0:00 /bin/sh /usr/local/mozilla/v1.7.13/mozilla >> 3197 ? S 0:00 \_ /bin/sh /usr/local/mozilla/v1.7.13/run-mozilla.sh >> /usr/local/mozilla/v1.7.13/mozilla-bin >> 3202 ? Sl 9:39 \_ /usr/local/mozilla/v1.7.13/mozilla-bin > >Sure looks like 3202 is right. >The Sl means there are threads, but all memory for all threads should be > accounted for in the process memory. I did "gcore" on 3195 and 3197 and both dumps were 5528 bytes. (Nothing much to see there.) > >> >> That's all. I'm guessing that the first one is the navigator, second is the main >> email program, and third is the email composer. That follows with my using one >> to spawn the next. All three are still open on the still frozen gui on F7 (just >> checked). >> >> Although I refer to the gui as being frozen, I don't think it is. With the mouse >> pointer still following the motions of the mouse, and the clock still ticking >> away, it makes me think that something else is confusing it. Alt-Tab does >> nothing, nor does Alt-Esc, Ctrl-Alt-Del, and not even mouse pointer context. >> Even Ctrl-Alt-[->] (right arrow) doesn't move to Workspace 2, but Ctrl-Alt-Fx >> switches to the appropriate console. Maybe the mouse got confused. It did >> freeze up during the midst of an attempt to drag an icon from one place to >> another. I'm just not quite ready to try anything that could cause Mozilla to >> self-destruct until I've made an acceptable effort to get my email. > >Yeah, I've seen this too, I don't really know a precise explanation of >what's going on in such a case. In some past cases, some of the desktop >features are also affected: panels go away or get blanked, application >windows get frozen in partially rendered state, ... This is not the first time this has happened while trying to drag an icon from one place to another. I'd *like* to know how to get it unstuck. This means learning nitty gritty stuff about linux. And most of the time, I'm at the mercy of the generosity of the gurus of the list. > >Clearly X has stopped working properly. Maybe the "event dispatching" >system is broke, or maybe some non-interruptible component failed to >"yield" or return. Is there a way to find out what froze and how to unfreeze it? I know *precisely* what I was in the midst of doing when it happened. *And* I know that this has happened before, I think the same exact way. > >Is a puzzlement! :-( > >> >>> How are you searching. I might try >>> >>> grep "known string" core.3202 >>> and if that works >>> strings core.3202 | grep -30 "known string" >>> to get (say) 30 text lines before and after the match. >>> >>> >>> If you get this far, checkout greps -B and -A options >>> .. grep -B10 -A50 .. >>> (or similar) might be useful. >> >> I launch vim with that file, and do "/knownWord". Got nuthen. > >Although I would still prefer something like > > strings core.3202 | less >or > strings core.3202 > core.strings > vim core.strings > >.. Vim _is_ pretty good at reading binary files, so I would expect what >you did to be ok. Do you see anything interesting? Did you try searching >for other (maybe shorter) strings you think might be there? > >The benefit of going through the strings operation, is that what results >is pure text, one line per contiguous sequence of text. The minimum >sequence of contiguous "printable" characters can be specified (see man >strings) - I think the default is 4 bytes. Anyway, if the non-ASCII >stuff is filtered out, the rest is easier to visually browse. I did "strings core.3202 > strings.core.3202" and then browsed through it. Nothing much worth noting except possibly some references to UTF8. Is it possible that the text is encoded in UTF8 and thereby appearing to be binary? > >I suppose if you do not discover any trace of the message you were >composing, there might be an explanation that some child process that >held that memory aborted (before X froze up), and it's memory got >unmapped and hence is not now accessible. I just fired up startx as a different user on :2 so that I could start up Mozilla. (I know the lock file prevents *one* user from having two instances at once, but not two users.) I started up Mozilla, then Mail, then Email composer, and entered some dummy text. Then I did gcore on that PID and then searched for the strings I had entered. Nada. So, either the text is in UTF8 or somesuch and not appearing as ASCII, or Mozilla has a scratchpad somewhere that is not immediately apparent. > >Back in the "oughta" world, it seems that there should be a way to >browse or dump the entire contents of RAM, but I don't know how. Once >upon a time (kernel 2.4, I think), you used to be able to read >/proc/kcore and see the active RAM content, but it no longer seems to >work that way. > >I have to ask: is the information you are looking for so valuable to >justify this amount of strain? I'm thinking that email you were >composing is a product of your own thinking, so even if it's annoying, >you should be able to repeat the mental process and get something >more-or-less equal to what was lost? It started off valuable (to me) because I was particularly pleased with the wording. Sometimes when the wording is that pleasing, reconstruction just never quite measures up. But, after a day or two, it became a growing curiosity to get under the hood. I mean, if I take the time now to learn how to get at the text that seems to be out of reach, later, when the info is really really important, I'll have a head start. For now, I'd rather find out how to get the frozen unfrozen without losing the running programs. (Though nobody responded to that question.) But I think that your comment may be better (in a way) where you said it would be good for a program to have a way to be passed a signal to save its contents. Unfortunately, there's a lot of programs out there that may or may not get that feature added. And it would still be nice to know how to get to the info. Perhaps if KDE or Gnome were to implement the sending of a "save" signal, maybe if the programs that run there start getting that signal, it might prompt the various programs' maintainers to take advantage of it. Out of curiosity, what kinds of signals are currently sent to running programs? > >I've lost multiple-page stuff on a number of occasions due to computer >failure coupled with my neglect of periodic save operations. Although it >sure is exasperating, I _have_ managed to continue my life :-[ .. such >as it is. ;-) Yeah, my life won't end without that email. It's an annoyance and a bump in the road. But nothing severe. > >Or, maybe it's more a matter of curiosity? If so, it might be worth >further pursuit (maybe with shorter, pointed questions): Why don't >programs have a capability to save work-in-progress upon receipt of some >system signal? Is there a way to dump an entire RAM image? Is there a class on how to think in "shorter, pointed questions"? ;> > >Regards, >..jim Again, thanks jim. ---- Msg sent via CWNet - http://www.cwnet.com/ -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
