Re: with-package-iterator

[Sam S. <sd...@gn...>]

Vladimir Tzankov wrote:
> When symbol is interned/uninterned - only symtabs change. On rehashes
> new symtab is created - so all threads that have gained access to old
> symtab will traverse it successfully (but will not see changes).
> Problem (not sure that it's the only one) with previous code was that
> sometimes the symtab stored in the state and its size may not be in
> sync (search1 - it was possible after getting the symtab and before
> the size -  another thread to cause rehash).

this functional (as in "no structure modifications") approach is convenient for 
MT, no doubt, but allocating a new symtab on each addition would degrade 
performance severely.


> Oops. I've been experimenting and left it by mistake - sorry.

it happens. :-)
because of this, I always at least try to proofread my own patches before 
committing ("=" in *vc* buffer before C-x v v).