[PATCH] patch for lin need_source function

[Jasper Huijsmans]

Auke Kok wrote:
> Jasper Huijsmans wrote:
> 
>> Auke,
>>
>> This seems to fix the problem with limitless lin's being spawned.
>> Doesn't seem to break anything else.
>>
>> The patch removes the subshell (why was it needed?) and adds some code
>> to make the list unique (uniq needs a sorted list, so won't always
>> work). Some other thing were changed as well.
> 
> 
> 
> I'm having difficulties understanding the patch:
> 

Me too, I'm afraid :/ It's not at all obvious to me why my version works 
and yours doesn't.

> 1) the subshell indeed is unneeded
> 

...but not the real problem, of course. Removing the subshell stops the 
spawning of 1000s of lin processes, but the loop is still infinite.

> 2) the uniq call without sort was deliberate, as the function returns 
> items based on build order. During testing I found that some items were 
> returned repeatedly, hence the uniq call. adding a 'sort' would perhaps 
> reduce the number of duplicates, but it also would distort the order.
> 

Ok, I think I noticed that. I just thought that it is possible to have 
different dependencies of a module depend on the same module and then 
create duplicates. (confusing sentence, hope it made sense).

> 3) I haven't had time enough to check out how it really works now, maybe 
> because also I don't really yet see how you apparently solved the probel 
> ;^)
> 

It's not obvious to me either. One difference is that I use only one 
echo statement to return from the need_source, because I catch the 
recursive calls. Seems to fix the duplicated too; could that be the reason?

Also I only run need_source when DEP is not installed. This saves a lot 
of needless recursion.

> I'm considering dumping the "new" method and going back to the old style 
> download (ie spool al immediately required modules on start, download 
> secondary requirements when needed).
> 

Well, it really does seem to work now :)

	Jasper