| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
| |
Follow-up to commit 3993: deep-copy needs to also ignore screen and
console objects. Basically channels and any fake objects passed into
tests. (The real ones will be null, which will work fine thanks to the
bugfix of commit 3996.)
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fully isolate routines from their arguments.
I still need exceptions for containers that are *designed* to be
shared between routines. The primary such case is channels; we need some
way to share them between routines, and if we deep-copy them that
defeats their entire purpose.
A milder case is the use of fake file-systems in tests, though that's a
hint that there'll be more of these as the OS gets more fleshed out. The
pattern seems to be that we need to not deep-copy containers that
contain lock fields, and so their operations internally do their own
locking.
We may have to stop hard-coding the list of exceptions and allow people
to define new ones. Perhaps don't deep-copy any container with metadata
of 'shared', and then ensure that get-location is only ever called on
shared containers.
This still isn't absolutely ironclad. People can now store something
into a channel and then pass it into a routine to share arbitrary data.
But perhaps the goal isn't to be ironclad, just to avoid easy mistakes.
I'd still want an automated check for this, though. Some way to
highlight it as an unsafe pattern.
This completes step 1 in the plan of commit 3992 for making
continuations safe.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Plan:
1. Fix a hole where addresses are shared between routines when passed in
as arguments to `start-running`.
2. Switch to a new approach to refcount management: instead of updating
refcounts when writing products of instructions by default, increment
refcounts inside instructions by default and decrement refcounts in
caller. More details in future when I actually implement this.
3. Now we shouldn't need a distinction between `new-default-space` and
`local-scope`, and all functions can simply decrement refcounts of
their default-space, consistently handling any refcounts in the space.
At this point if all goes well, continuations should be safe!
This commit is just preparation for step 1.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
They're back after a long hiatus: commit 2295 in Oct 2015.
I'm not convinced anymore that this is actually a correct implementation
of continuations. Issues on at least two fronts:
a) These aren't safe yet. Since continuations can be called multiple
times, we need to disable reclamation of locals inside a continuation.
There may be other type- or memory-safety issues. However, delimited
continuations at least seem possible to make safe. Undelimited
continuations (call/cc) though are permanently out.
b) They may not actually be as powerful as delimited continuations.
Let's see if I can build 'yield' out of these primitives.
|
| | |
|
| |
|
|
| |
This regression was (also) introduced by commit 3902 in June.
|
| | |
|
| |
|
|
|
|
| |
This regression was introduced by commit 3902 in June.
Making this commit clean took the last 4 commits of reorganizing.
|
| |
|
|
|
| |
It's always been ugly that I referred to a later layer/feature in a
label name.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
We never ended up using this, and all it has provided is a hole in the
type system.
Java's lesson is that globals are not essential. Any program that uses
globals can be rewritten to do without them. Just thread a world object
through the program, saving global state in it, and extracting relevant
bits of global state to pass into functions lower down the call chain.
Thanks Andrew Owen for the conversation that led me to focus on this
issue.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Instead of setup() and teardown() we'll just use a reset() function from
now on, which will bring the machine back to a good state before each
test or run, and also before exit (to avoid memory leaks).
|
| |
|
|
|
| |
I'm in the process of making it more self-contained so I can use it in
another project.
|
| |
|
|
|
|
|
|
| |
Narrow the scope of implicit type conversions. Now only numbers can be
freely converted to from other scalars (booleans, characters). We want
in particular to make this an error:
x:character <- new [abc]
|
| |
|
|
|
| |
Yet another regression, this time cascading from commit 3953. My
scenario wasn't actually testing what I thought it was testing.
|
| |
|
|
|
|
|
|
|
| |
Expand the steps in `update_html` and try to process each set of files
separately so we can see commonality.
The eventual goal is a script that can selectively process a subset of
files. But this is a good first step: I can at least easily comment out
different subsets.
|
| |
|
|
|
| |
Don't refresh entire sandbox side when toggling the expected result for
a single sandbox.
|
| |
|
|
|
|
|
| |
Don't unnecessarily write sandboxes to disk on F4.
This seems to save almost 20% time when processing a large lesson
directory with 36 sandboxes.
|
| |
|
|
|
|
|
|
| |
Improvement on fix 3957: rather than put a band-aid over a slow
operation, eliminate the slowdown entirely.
In this case it turns out we're unnecessarily saving files to disk when
they could never be modified. Are we doing this on F4 as well?!
|
| |
|
|
|
| |
Marking sandbox responses as expected can take a long time if there are
many sandboxes. Indicate when a click is being worked on.
|
| | |
|
| |
|
|
|
| |
Move a scenario which is after commit 3954 applicable to both editors,
not just the recipe side.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As a blanket rule, down-arrow now stops scrolling once the bottom margin
comes on screen.
Now that we have page-wise scrolling with ctrl-f/b and line-wise
scrolling with ctrl-s/x, we don't need to conflate scroll positioning
with the arrow keys. And as a result, early students no longer have to
struggle with accidentally scrolling part of the sandbox off the screen
when there's tons of empty space available.
`move-to-next-line` is still super messy and will need further
rethinking, but this commit simplifies the codebase as a whole by
eliminating a couple of historical accidents:
a) We only introduced scrolling past the bottom of the screen to allow
more sandboxes to come into view before we had scrolling for the
sandbox side.
b) We undid scrolling past the bottom in just the recipe side to allow
errors to come into view.
Since these historical details are now irrelevant, we no longer need
separate logic for the recipe and sandbox sides, and we don't need to
keep track of the recipe-bottom separate from the bottom margin of
arbitrary editors.
|
| |
|
|
|
|
|
|
|
| |
Redo cursor-down to fix a bug.
This should be the final bug I find as part of my audit of everywhere we
compare characters to newlines in layer 3 of the edit/ app.
Pretty messy implementation, but we'll clean it up now.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Fix a regression caused by commit 3919.
Thanks Juan Crispin Hernandez for running into this.
|
| | |
|
| | |
|