You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The job still gets held. Same held reason. But I get a different stderr.
can't open file '0': [Error 2] No such file or directory
which is actually a bug all on its own; it wasn't intentional that map options from settings can bypass the reserved descriptors check (executable is reserved), but that is indeed exactly what happens, because that check only occurs when constructing the MapOptions, but the settings map options are merged in during create_submit_object_and_itemdata() via get_base_descriptors(). See
Not quite sure what the logic was supposed to be here, but it's wrong. The intent is that map options in settings should be injected into all MapOptions, so we should probably just do that in the constructor. Might be worth another look at how MapOptions are built up in the first place, because they're pretty weird under the hood.
The text was updated successfully, but these errors were encountered:
The core problem here is that the merge is happening at the wrong location, and therefore does not serve the intended purpose. Need to think hard about what the intent of default options from settings is, which will indicate where they should be merged, especially given that they can't really be Python types.
JoshKarpel
changed the title
Map options from settings bypass reserved submit descriptor check
Map options from settings are merged incorrectly
Apr 27, 2020
In #198 , @keith6014 noted that
which is actually a bug all on its own; it wasn't intentional that map options from settings can bypass the reserved descriptors check (
executable
is reserved), but that is indeed exactly what happens, because that check only occurs when constructing theMapOptions
, but the settings map options are merged in duringcreate_submit_object_and_itemdata()
viaget_base_descriptors()
. Seehtmap/htmap/options.py
Line 287 in bb20219
Not quite sure what the logic was supposed to be here, but it's wrong. The intent is that map options in settings should be injected into all
MapOptions
, so we should probably just do that in the constructor. Might be worth another look at howMapOptions
are built up in the first place, because they're pretty weird under the hood.The text was updated successfully, but these errors were encountered: