| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Comments:
<0> D-side: using mr double-bladed-battleaxe?
<1> in the donotational way of the word
<1> ooooooogling
<0> mighty is your google-fu, young K_F
<2> hehe
<3> garbage in, google out
<1> okay, less vodka, less irc, more sleep
<3> wuss
<0> yeah, bed time here too
<1> Ka-bar: purgamentum init, exit purgamentum
<0> got someone to whap around the head tomorrow and suggest he read lots of rfcs so he can hold up his own end of an intelligent conversation...
<3> illegitimi non carborundum ;)
<0> "its not a proper mime attachment because it doesnt' have a flle name"
<3> HA
<1> Ka-bar: are they hot chicks?
<0> bloody things dont' even have to have a content type if they don't want to
<2> mime attachments are the work of satan.
<0> no, mime multipart/related is the work of satan
<0> mime is just the work of a minor devil :)
<3> no, rfc1759 is the work of satan
<2> I'll have to read it .... brb
<0> didn't postal draft much of the mime stuff?
<4> and rfc2549 is ... who? Bill Clinton?
<3> no, rfc2549 is actually entertaining. rfc1759 was serious and I'm now actually ashamed that my name is in it.
<3> Of course... it did get me my current job, so I guess I can't complain too much...
<5> help. if i want to mount /etc from a different filesystem each time linux boots, where/how can i do it?
<0> where/how isn't as interesting as "WTF???"
<6> night folks
<7> poof
<4> Ka-bar: huh? Which one are you? Xerox? TI? Lexmark?
<3> Underscore
<2> DaveHowe: you are correct, I didn't give Mr. Google the proper level of interrogation. I hang my head in shame, while blaming the beer.
<3> you can only blame the beer if it was really good beer.
<2> Ka-bar: Now I'm truly owned, I'm drinking cheap nasty g***y ****.
<2> Ahhhh ... the good stuff.
<2> Skol!
<4> Ka-bar: huh? Undersore?
<2> Ka-bar: turn off the phone!
<3> Viking667: I wasn't one of the principal authors of the RFC but I was a major contributor to it and attended most of the meetings of the group. My name is in the list of contributors at the end. I was working for a software company called Underscore at the time. My experience doing SNMP **** is what got me my current job.
<4> ahhhh.
<3> putting SNMP in another type of appliance...
<4> Ka-bar: hhhmmmm. Kevin?
<3> Viking667: *nod*
<8> ewww snmp
<9>
<4> heh. I'm wary of turning off rpc-based stuff in Solaris, because I don't know what actually DEPENDS upon its availability for execution...
<4> I'm not exactly willing to turn it off, and then not be able to log in again....
<2> I use mrtg at my workplace, which relies on snmp. We have a bit of a political thing where whenever mrtg spikes and it;s down to a reporting error, people go crazy and demand fault tickets be raised. I dont care, I'm ignoring it, it's all very silly.
<2> not snmp or mrtg ... my workplace.
<7> Liandrin!
<3> Viking667: NFS depends on RPC...
<7> V667
<4> Ka-bar: heh. I don't use NFS here, but my home volume is mounted using autofs... that's about as close as I get.
<4> hi Nanuq
<4> how come I don't get the exclamation marks? Is Liandrin wonderful or something?
<8> Viking667, of course :P
<10> it's good champony ?
<8> does Solaris still use rpc.authd?
<11> hey.. does anyone know how (or if it is possible) to check the reference count of a shared library that has been dlopen()'ed and dlclosed()'ed possibly a few times?
<11> and for extra bonus points... does the linux memory manager automatically unmap the library from the memory when the reference count reaches zero? or does it wait until it actually needs the memory back for other stuff
<11> :)
<11> (basically... in my application, i'm dlopening a library once, and then dlclosing it once... and after the close, it is still mapped in memory, seemingly forever... and i dont know why)
<11> and this is on an imbedded system that cannot do any swapping
<3> ryan1: 'man dlclose'
<3> it says that "If the reference count drops to zero and no other loaded libraries use symbols in it, then the dynamic library is unloaded."0
<11> been there a million times... man dlclose says it decrements the reference count... no **** :)... it doesnt say anything about linux memory management/replacement
<11> yeah, but it doesnt say if it is unloaded immediately or at some later point
<11> nor does it say how i can verify what the ref count is :(
<8> "If the reference count drops to zero and no other loaded libaries use symbols in it, then the dynamic library is unloaded."
<11> and i spent a few hours trying to weed thru the source of glibc with no luck
<12> hello everybody
<13> to me it says it is closed the moment when ref count is zero and no symbols are used anymore
<8> that sounds fairly conclusive
<11> but is "Unloading the dynamic library" == "unmapping the library from memory" ?
<12> Is there someine who can help me with a boot problem? I have a laptop with a livecd knoppix 5.0 and the damn thing hangs when it tries to mount the cdrom. Older versions of knoppix work ok. I have no idea.
<11> because that's not what i'm observing
<8> of course, if your library loads other libraries, they may be construed as keeping your library pinned
<14> Is it easily possible to add files to a .img file?
<3> I would guess that it happens at some later point. Deallocation is lazy.
<11> it does load other libraries implicitly, but not explicitly with dlopen
<8> ryan1, then they may well be pinning your library down.
<3> and I do think that "unloading" means freeing the memory
<12> Are there any known problems with CD and laptops. It's just weird that older versions work
<3> zanshin: maybe bad media
<14> I'm trying to inject a file into an image.
<3> TradeMark: huh?
<12> nope it boots on my desktop machine
<11> do you know what kernel level system calls the glibc dlopen/dlclose actually call? i ***ume mmap/munmap
<14> Insert file x into y.img
<14> Is it possible?
<3> ryan1: it talks to ld.so
<3> TradeMark: I have no idea what you are talking about. In what way? Be specific.
<8> ryan1, have you tried adding a _fini routine?
<3> TradeMark: be *very* specific. We are not mind readers
<11> no... what is _fini?
<8> ryan1, man dlclose
<14> I need to get a file into QEMU, the easiest way I can see to do it is to create an img file containing the file in question, along with some boot files.
<11> oh sweet
<11> that'll let me know if it is closed
<11> or if it is just waiting for linux to free the memory later
<11> thanks
<8> TradeMark, it depends on the kind of image file. if it's an ISO-9660 filesystem image (i.e., a CD), no
<3> ryan1: are any other programs using the same library that you are 'dlopen'ing?
<11> no just one program
<8> TradeMark, otherwise, you can probably mount the filesystem image via the loopback device (mount -oloop filename /mount/point) and futz with it
<14> Being very specific... I've installed W2K from a SP3 cd and it's having issues with msgina.dll. I have a newer version of the file, but I need to boot into a console somehow and replace the file.
<14> Hmm ok I'll give that a shot.
<3> W2K?
<8> TradeMark, of course, if it's NTFS, you're probably screwed, since read-write support is very flaky in Linux for that ;)
<14> FAT32
<8> ah, you're in the clear, then.
<11> Liandrin - thanks for letting me know about _fini. much appreciated.
<14> Bah.
<14> Liandrin: Should I be able to interact with it directly? I get "not enough space" when trying to copy the file into the image.
<3> ryan1: also look in /proc/<pid> while the process is running to see what is going on
<11> yeah
<11> the reason i know it is still mapped in memory
<11> is because i still see the library listed in /proc/pid/maps
<3> hmmmm
<11> i still dont know what the 3rd column of stuff in 'maps' is though.. i ***ume some sort of bit mask
<3> wellllllll..... you could try running some other processes to incrementally hog memory and see if it eventually gets unloaded.
<11> yeah... hard to tell though when i run my app on simulation, when there is swap memory available
<11> and i dont have access to the actual target hardware platform yet :(
<3> ryan1: 'man proc' describes what's listed in /proc/pid/maps
<3> ryan1: just turn swap off. ;)
<11> ahh offset
<11> heh i might be able to get away with that... of course if i go too far my workstation dies :)
<3> ryan1: I've also done some embedded work and have occasionally had to play tricks like that on a dev system to track down a problem.
<8> TradeMark, uh
<11> ideally, i'd like to see some FOSS that makes use of loading/unloading plugin shared libarires to see a sample...
<8> TradeMark, have you checked that there is actually enough space available in the image? ;)
<11> of course, i dont know of any offhand... might be why firefox requries restarting the browser when you remove plugins ;)
<11> because it's a PITA :)
<14> Liandrin: I think I've got it sorted, I'm copying out the files and making a new image.
<8> ryan1, the overwhelming majority of things that play libdl games don't really care if/when the loaded objects get discarded
<11> yeah i figured as much..
<8> ryan1, as for Firefox, its plugins can entwine themselves into any and all aspects of the browser, and aren't required (afaik) to implement a "get lost" function
<11> most apps load plugins... and are there until the process is terminated
<11> of course, in my application, this process is not supposed to ever terminate
<3> does the process fork() or use threads?
<11> no and no
Return to
#linux
or
Go to some related
logs:
#chatzone
me quiere cojer
kacedonija
#linux
#chatzone
#MissKitten
ublas portability
mysql tables_in_
#c
#chatzone
|
|