Leo-

Op 10 nov 11:35 schreef Leo Famulari:
> On Fri, Nov 10, 2017 at 08:26:12AM +0100, Marco van Hulten wrote:
> > kodi@watson ~$ time guix pull --cores=1  
> 
> [...]
> 
> > compiling...	 75.4% of 647 filesbuilding of `/gnu/store/gk5chb0dwpsq3na7b4gn1hd7h0h2b63h-guix-latest.drv' timed out after 3600 seconds of silence  
> 
> [...]
> 
> > real	609m30.245s
> > user	0m3.125s
> > sys	0m0.875s  
> 
> This is terrible, but you can work around it by passing a large value to
> the --max-silent-time "common build option": [...]

Hmm, but the time-out is now after 1 hour of silence, and the whole
process takes over 10 hours before crashing.  This option may be useful
if I set a very short time, though it remains to be seen if this ends
guix quickly.  I will play around with it.

Right now a `guix pull' results in a `dispatch-exception' in procedure
`struct_vtable: Wrong type argument in position 1 (expecting struct):
#<pointer 0x0>'.  It seems that GuixSD (or components) is strongly in
flux right now, so I will try again later and do a proper bug report if
problems persist.

> Out of curiosity, what kind of machine are you using? A full hour with
> no output at all indicates that something is happening very slowly! Is
> it swapping?

I'd like to test if it is swapping.  I have 2 GiB of RAM in a 2-core
machine (Intel Core 2 Duo).  I did notice that swap was not enabled yet
on my system, and that (during last `guix pull') around 95% of RAM was
used.

> > - it took over 10 h to give me back control, whereas this used to be
> >   a bit over 2 h in previous tries;  
> 
> Do you mean that the computer becomes unresponsive?

Yes.

> > What is Guix doing between 75.2 and 75.4 %?  
> 
> I don't know exactly, but during `guix pull` Guix is built from source
> with the Guile compiler.
> 
> There are some bugs in the current version of the Guile compiler that
> cause it to require way more memory than expected. This is terrible on
> memory-constrained systems because it forces the use of swap, which is
> typically super slow. Hopefully we can deploy a fix soon.

Is 2 GiB considered "memory-constrained"?

-Marco