| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
| |
Always show instruction before any transforms in error messages.
This is likely going to make some errors unclear because they *need* to
show the original instruction. But if we don't have tests for those
situations did they ever really work?
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Another bug, this time in cycle detection, which had been present from
the start but hadn't actually been tested until now. Turns out it had
two bugs, one with the map not being passed into a recursive call, and a
more subtle one: I wasn't actually saving the right input address.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
Gaining confidence now in the fix of 3086. Basically I forgot that
size_of doesn't canonize its argument (because it needs to run at
transform-time, but canonize cares about the value of the reagent for
arrays, and value is only available at run-time). So I was always
returning the size of the address (1) as the size of its payload.
|
| | |
|
| |
|
|
|
|
|
|
|
| |
More tests, another bugfix. I still don't *really* understand why it
works, though..
The new tests aren't really done yet. No assertions, lots of debug info.
Just translated from sandboxes. deep_copy_stress_test_2 requires the
bugfix.
|
| |
|
|
|
|
| |
To my surprise, I might be done with deep-copy. Suddenly all the holes I
had for special cases have melted away. At least any new bugs will be
holes I'd never thought of.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Back to slow progress on deep-copy.
|
| |
|
|
| |
Support for containers without addresses.
|
|
|
Just simple non-address primitives so far.
|