Stream: brlcad

Topic: datums


view this post on Zulip Sean (Jul 23 2020 at 12:59):

Gah, how did datums get busted?!?!

view this post on Zulip Sean (Jul 23 2020 at 13:04):

looks like minortype is correct (44).. majortype is 255, though, and that's not right

view this post on Zulip Sean (Jul 23 2020 at 13:06):

yeah, it's supposed to be 1 (majortype BRLCAD)

view this post on Zulip starseeker (Jul 23 2020 at 13:20):

Busted in 7-30-2-1

view this post on Zulip starseeker (Jul 23 2020 at 13:20):

works in 7-26-0

view this post on Zulip Sean (Jul 23 2020 at 13:20):

looks like the datum_in code has never set the majortype but worked, so either extra validation checks were added or majortype setting was removed from someplace common to multipl types

view this post on Zulip starseeker (Jul 23 2020 at 13:20):

OK, you're deeper than I am - want me to add a regression test while you're hunting?

view this post on Zulip Sean (Jul 23 2020 at 13:20):

can you see what the majortype is in 26.0?

view this post on Zulip starseeker (Jul 23 2020 at 13:21):

what's a quick way to print that?

view this post on Zulip Sean (Jul 23 2020 at 13:21):

does the datum from 26 open in 30?

view this post on Zulip starseeker (Jul 23 2020 at 13:22):

Opens in trunk, actually - draws too

view this post on Zulip Sean (Jul 23 2020 at 13:23):

okay, so it's not extra validation then

view this post on Zulip Sean (Jul 23 2020 at 13:23):

just to confirm, you can try running: gex file.g

view this post on Zulip Sean (Jul 23 2020 at 13:23):

see if there's an unknown majortype listed (assuming you didn't create invalid and valid ones)

view this post on Zulip starseeker (Jul 23 2020 at 13:23):

===============================================
Objects found in BRL-CAD V5 database: datum.g
===============================================

Object DLI type/major/minor type: 1/ 0/ 0 name: (none)
Object DLI type/major/minor type: 0/ 2/ 0 name: _GLOBAL
Has one or more attributes.
Object DLI type/major/minor type: 0/ 1/ 44 name: datum.s
Has one or more attributes.

==========================================
Summary for BRL-CAD V5 database: datum.g
==========================================

Found 3 objects:
free space: 0
named : 2
other : 1

2 objects have one or more attributes.

Object DLI types (the main category: defined in H Flags):
0 (APPLICATION_DATA_OBJECT) : 2
1 (HEADER_OBJECT) : 1
Object major types:
0 (RESERVED) : 1
1 (BRLCAD) : 1
2 (ATTRIBUTE_ONLY) : 1
Object minor types:
0 (RESERVED) : 2
44 (unknown) : 1

Note: file read ended normally at EOF.

view this post on Zulip Sean (Jul 23 2020 at 13:24):

yeah that looks right

view this post on Zulip Sean (Jul 23 2020 at 13:24):

created prior to my fix creates:

view this post on Zulip Sean (Jul 23 2020 at 13:24):

Object major types:
0 (RESERVED) : 4
1 (BRLCAD) : 14
2 (ATTRIBUTE_ONLY) : 1
9 (BINARY_UNIF) : 1
255 (unknown) : 1

view this post on Zulip Sean (Jul 23 2020 at 13:25):

ah, right, there it is: Object DLI type/major/minor type: 0/ 1/ 44 name: datum.s

view this post on Zulip Sean (Jul 23 2020 at 13:26):

vs Object DLI type/major/minor type: 0/255/ 44 name: datum

view this post on Zulip starseeker (Jul 23 2020 at 13:27):

busted in rel-7-28-2

view this post on Zulip Sean (Jul 23 2020 at 13:27):

this is where having a database with all objects in the regression suite would have triggered a failure, caught it

view this post on Zulip Sean (Jul 23 2020 at 13:27):

.. @Thusal Ranawaka can you make sure you add datums to your list? :)

view this post on Zulip starseeker (Jul 23 2020 at 13:32):

rel-7-26-4 works

view this post on Zulip starseeker (Jul 23 2020 at 13:38):

@Sean Is r76443 the full fix? I'll have to start bisecting if you want me to ID which commit actually broke it

view this post on Zulip starseeker (Jul 23 2020 at 13:40):

May want to put out a patch release with that fix - trunk would be a heck of a jump just for that...

view this post on Zulip starseeker (Jul 23 2020 at 13:42):

Working as of r71000

view this post on Zulip Thusal Ranawaka (Jul 23 2020 at 13:51):

Sean said:

.. Thusal Ranawaka can you make sure you add datums to your list? :)

@Sean Roger that! :police:

view this post on Zulip starseeker (Jul 23 2020 at 14:27):

Found it - I did it, commit r71953

view this post on Zulip Sean (Jul 23 2020 at 15:12):

I don't think we need to bisect the cause

view this post on Zulip Sean (Jul 23 2020 at 15:12):

oh, I see you found it, great

view this post on Zulip Sean (Jul 23 2020 at 15:14):

so what that might imply is that other 'in' command primitives could be busted. we should check them since this reared its head

view this post on Zulip starseeker (Jul 23 2020 at 15:15):

/me agrees

view this post on Zulip Sean (Jul 23 2020 at 15:15):

most manually set, but I did see at least two other commits where the type was added back in, presumably to fix a released error

view this post on Zulip Sean (Jul 23 2020 at 15:16):

also implies a NEWS item will be in order for those types

view this post on Zulip starseeker (Jul 23 2020 at 15:16):

/me nods - I'm thinking a 7.30.10 patch release? Should be easy to backport the changes for this issue...


Last updated: Oct 09 2024 at 00:44 UTC