infinite recursion org-update-id-locations / org-entry-properties / org-entry-blocked-p / org-edna / org-id-find / org-id-update-locations

infinite recursion org-update-id-locations / org-entry-properties / org-entry-blocked-p / org-edna / org-id-find / org-id-update-locations

[Justin Vallon]

I copy by org files between 2 machines.  Thus, my .org-id-locations file
gets stale when I archive on one machine, then sync the org files.

I also use org-edna.

Sometimes, org-agenda will go into an infinite loop (recursion depth,
etc) when scanning.

I believe what is happening is:

- org-agenda / org-edna builds the agenda, and hits a "missing" id(X)
- org-update-id-locations starts scanning files/nodes to refresh
- org-cached-entry-get tries to get all properties
- org-entry-blocked-p is called
- org-edna then tries to parse the BLOCKER property
- the property mentions "id(Y)", which has not been found (yet)
- cache miss, rescan files

The X and Y could be different, but once id(Y) is missing and the
cache-miss happens during id-scanning, it will repeat.

I suppose it also might be sensitive to the order of files in the list,
since scanning the archive first would scan the older IDs, then the
active file would reference those older IDs.  However, I would imagine
there would be trouble if an "earlier" node in the active file
referenced an ID that appeared "later" in the file, since the file is
being scanned in-order.

I haven't tried this, but the simplest case to reproduce might be to
delete the .org-id-locations file, then feed it:

* Second
  BLOCKER: id(first)
* First
  ID: first

I can try to build a (simple) reproducible test case if this is unclear.

My work-around is to disable org-edna, run org-id-update-id-locations,
then reenable org-edna.

-- 
-Justin
JustinVallon@gmail.com

Re: infinite recursion org-update-id-locations / org-entry-properties / org-entry-blocked-p / org-edna / org-id-find / org-id-update-locations

[Justin Vallon]

I just tried using straight for org, and master (a1e5bee5c) has a
rewritten org-update-id-locations that uses regexps to find the IDs.  It
appears to have fixed the problem (for cases that I tried).

-Justin


On 7/30/20 4:21 PM, Justin Vallon wrote:
> I copy by org files between 2 machines.  Thus, my .org-id-locations file
> gets stale when I archive on one machine, then sync the org files.
> 
> I also use org-edna.
> 
> Sometimes, org-agenda will go into an infinite loop (recursion depth,
> etc) when scanning.
> 
> I believe what is happening is:
> 
> - org-agenda / org-edna builds the agenda, and hits a "missing" id(X)
> - org-update-id-locations starts scanning files/nodes to refresh
> - org-cached-entry-get tries to get all properties
> - org-entry-blocked-p is called
> - org-edna then tries to parse the BLOCKER property
> - the property mentions "id(Y)", which has not been found (yet)
> - cache miss, rescan files
> 
> The X and Y could be different, but once id(Y) is missing and the
> cache-miss happens during id-scanning, it will repeat.
> 
> I suppose it also might be sensitive to the order of files in the list,
> since scanning the archive first would scan the older IDs, then the
> active file would reference those older IDs.  However, I would imagine
> there would be trouble if an "earlier" node in the active file
> referenced an ID that appeared "later" in the file, since the file is
> being scanned in-order.
> 
> I haven't tried this, but the simplest case to reproduce might be to
> delete the .org-id-locations file, then feed it:
> 
> * Second
>   BLOCKER: id(first)
> * First
>   ID: first
> 
> I can try to build a (simple) reproducible test case if this is unclear.
> 
> My work-around is to disable org-edna, run org-id-update-id-locations,
> then reenable org-edna.
> 


-- 
-Justin
JustinVallon@gmail.com