Failure when Re-running Tests

[brodieG]

# within validate..
unitize_dir()
# poke around a bit, finally, select 5 for validate.R
N # a bunch of times
devtools::install()
R
Y

and at this point:

| Re-run unitizer ([Y]es, [P]rev, [B]rowse, [U]nreviewed, [R]erun, f[O]rce)?

unitizer> Y

Error in slot(z, i)[[id]] : subscript out of bounds
| Unexpectedly exited before storing unitizer; if you were reviewing a unitizer 
| changes to that unitizer were not saved.  Note that any unitizers you *
| completed* review of should have been saved.

[brodieG]

Actually, happened with an even simpler scenario:

> unitize_dir()

| Summary of files in common directory 'unitizer':                              
| 
|                      Pass Fail
| *1.          eval.R    19    9
|  2.          misc.R     9    -
| *3.         parse.R    40    9
| *4. validate.args.R    15   13
| *5.      validate.R    41   20
| ..............................
|                       124   51
| Legend:
| * `unitizer` requires review

| Type number of unitizer to review, 'A' to review all that require review

unitizer> 5
+------------------------------------------------------------------------------+
| unitizer for: unitizer/validate.R                                            |
+------------------------------------------------------------------------------+

                       Pass Fail
          Tokens Pass    22    -
          Tokens Fail    16   14
 Compound Expressions     3    3
   Other Return Modes     -    3
................................
                         41   20

= Tokens Fail ==================================================================

- Failed -----------------------------------------------------------------------

| The 14 tests in this section failed because the new evaluations do not match 
| the reference values from the store. Overwrite with new results ([Y]es, [N]o, 
| [P]rev, [B]rowse, [R]erun, [Q]uit, [H]elp)?

> validate(INT.1, 1.2)
Error in validate(INT.1, 1.2) : 
  Argument `current` should `current` should be type "integer-like" (is "double"
)

| Conditions mismatch: 

< .REF$conditions                                                               
> .NEW$conditions                                                               
@@ 1,3 / 1,3 @@                                                                 
  Condition list with 1 condition:                                              
< 1. Error in validate(INT.1, 1.2) : Argument `current` should be type "integer-
<    like" (is "double")                                                        
> 1. Error in validate(INT.1, 1.2) : Argument `current` should `current` should 
>    be type "integer-like" (is "double")                                       

| State mismatch; see `.DIFF$state` for details.

unitizer> N

> validate(INT.1, 1:2)
Error in validate(INT.1, 1:2) : 
  Argument `current` should `current` should be length 1 (is 2)

| Conditions mismatch: 

< .REF$conditions                                                               
> .NEW$conditions                                                               
@@ 1,3 / 1,3 @@                                                                 
  Condition list with 1 condition:                                              
< 1. Error in validate(INT.1, 1:2) : Argument `current` should be length 1 (is  
<    2)                                                                         
> 1. Error in validate(INT.1, 1:2) : Argument `current` should `current` should 
>    be length 1 (is 2)                                                         

| State mismatch; see `.DIFF$state` for details.

unitizer> Q

| No changes recorded.
Error in slot(z, i)[[id]] : subscript out of bounds
| Unexpectedly exited before storing unitizer; if you were reviewing a unitizer 
| changes to that unitizer were not saved.  Note that any unitizers you *
| completed* review of should have been saved.

[brodieG]

Even simpler recreation with:

> unitize('validate.R')

| Inferred test file location: tests/unitizer/validate.R

+------------------------------------------------------------------------------+
| unitizer for: unitizer/validate.R                                            |
+------------------------------------------------------------------------------+

                       Pass Fail  New
          Tokens Pass    22    -    -
          Tokens Fail    30    -    -
 Compound Expressions     4    2    3
   Other Return Modes     -    3    -
     Multi-line Stuff     5    -    -
.....................................
                         61    5    3

= Compound Expressions =========================================================

- Failed -----------------------------------------------------------------------

| The 2 tests in this section failed because the new evaluations do not match 
| the reference values from the store. Overwrite with new results ([Y]es, [N]o, 
| [P]rev, [B]rowse, [R]erun, [Q]uit, [H]elp)?

# Fail

> validate(INT.1 || NULL, 1.4)
Error in validate(INT.1 || NULL, 1.4) : 
  `1.4` should be type "integer-like", or "NULL" (is "double")

... snip

unitizer> Q

| No changes recorded.
Error in slot(z, i)[[id]] : subscript out of bounds
| Unexpectedly exited before storing unitizer; tests were not saved or changed.

The problem seems to be that the stored reference state is empty (i.e. corrupted), so in this part of the code state.merged <- mergeStates(items.ref, x@state.new, x@state.ref) in R/browse.R x@state.ref doesn't have anything in it even though items.ref references specific items. The problem is in the actual unitizer rds as shown by:

str(readRDS('tests/unitizer/validate.unitizer/data.rds')@state.ref)

Loading the rds from commit c3742e9 shows a populated @state.ref, so something went terribly wrong later. Need to figure out why @state.ref is not getting populated, and whether it is a fluke issue with this unitizer or a more systematic problem that was introduced.