coreutils 5.0 install hang

[Simon Brenner]

Auke Kok wrote:

>On Wed, 2003-11-05 at 22:11, Terry Chan wrote:
>  
>
>>Short answer: Don't do that.                                                                                                        
>>
>>Please stop using tmpfs for your /tmp.     
>>    
>>
>
>urr....
>
>does he have tmpfs on for /tmp or for lunar compiles
>(/usr/src/$MODULE-$VERSION, or the $TMPFS option in the lunar options
>menu?)
>
>if he has only tmpfs on /tmp, and reserved 256mb for that... he just has
>a full /tmp (and should not complain in here about that, unless it's
>still the default on the latest ISO, and he happened to install that).
>  
>
/tmp was *empty*, until after the configure test filled it up with an 
installwatch log. The /usr/src/* directory is as it should be, ~30megs 
or so I think it was... And both /tmp and /usr/src/* use tmpfs, which 
has worked very well for me (except that you need a lot of swap to build 
xfree or mozilla or gcc)

>either way I'd like to know if this was just a full /tmp or a full
>/usr/src/{module on TMPFS}...
>
>sofar
>
>  
>
>>Do not enable tmpfs for your compiles either.  Instead use the -pipe option                                                         
>>in the "lunar optimize" menus.                                                                                                      
>>
>>256MB is not nearly large enough to compile anything of consequence.  Back when                                                     
>>some people used tmpfs for compiling, mozilla could chew up 1.5Gig of tmpfs.                                                        
>>
>>Terry Chan
>>--------------------------------------------------------------
>>On Wed, Nov 05, 2003 at 03:22:21PM +0100, Simon Brenner wrote:
>>    
>>
>>>when doing a lin coreutils (or lunar renew - problem seems unrelated to 
>>>how the installation is started), the configure script hangs at the 
>>>getcwd tests.. If I leave the installation running for a few minutes and 
>>>then hit Ctrl-C (or kill the conftest process), I get some error 
>>>messages about "Could not delete ...: No space left on device" (seems 
>>>that tmpfs is a little crooked in some places..), and watching df when 
>>>the installation is still running I could see that all the 256meg quota 
>>>for /tmp is filled, by a 256meg installwatch log.
>>>
>>>I inspected the tail of the installwatch log, and it seems that the test 
>>>program has gotten stuck in a loop trying to rmdir(..) folders. The test 
>>>program creates lots and lots of nested subdirectories named confdir3, 
>>>to test the behaviour of getcwd, then backs out and rmdirs each 
>>>subfolder one after one: /usr/src/coreutils/confdir3/confdir3/....., 
>>>/usr/src/corutils/confdir3, /usr/src/confdir3, /usr/confdir3, /confdir3, 
>>>and it seems to get stuck on that last folder, trying to delete it over 
>>>and over...
>>>
>>>      
>>>
>>_______________________________________________
>>lunar mailing list
>>lunar at lunar-linux.org
>>http://dbguin.lunar-linux.org/mailman/listinfo/lunar
>>    
>>