for #10993: cancellation sensitivity for buffered writer#11052

Draft

problame wants to merge 22 commits intoproblame/enspc-must-not-use-after-returned-errorfrom problame/enspc-must-not-use-after-returned-error--cancellation-sensitivity

+187-93

Commits on Feb 28, 2025

Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, explore alternatives
problame
committed
Revert "Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, explore a…
problame
committed
alternative approach to cancellation where we wait for interest in the result to go away
problame
committed
alternative approach: wait for peer shutdown
problame
committed
alternative approach: some WIP testing
problame
committed
Revert "alternative approach: some WIP testing"
problame
committed
Revert "alternative approach: wait for peer shutdown"
problame
committed
Revert "alternative approach to cancellation where we wait for interest in the result to go away"
problame
committed
WIP explore what happens if we make the fush loop infallible
problame
committed
WIP onto something
problame
committed
Revert "WIP onto something"
problame
committed
idea: inline the FlushHandle::flush() method and reorder to make mutable() always present; nice in theory but causes lost writes if flush() is retried
problame
committed
Revert "idea: inline the FlushHandle::flush() method and reorder to make mutable() always present; nice in theory but causes lost writes if flush() is retried"
problame
committed
Revert "WIP explore what happens if we make the fush loop infallible"
problame
committed
re-cherry-pick: Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, e…
problame
committed
idea: remain read-available after Timeline::cancel has been cancelled; walreceiver shouldn't be retrying any writes, and will cause a panic if it does
problame
committed
WIP
problame
committed
WIP
problame
committed
it compiles
problame
committed
all rust tests pass
problame
committed
improve docs
problame
committed
linux
problame
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

for #10993: cancellation sensitivity for buffered writer#11052

Commits on Feb 28, 2025

Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, explore alternatives

Revert "Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, explore a…

alternative approach to cancellation where we wait for interest in the result to go away

alternative approach: wait for peer shutdown

alternative approach: some WIP testing

Revert "alternative approach: some WIP testing"

Revert "alternative approach: wait for peer shutdown"

Revert "alternative approach to cancellation where we wait for interest in the result to go away"

WIP explore what happens if we make the fush loop infallible

WIP onto something

Revert "WIP onto something"

idea: inline the FlushHandle::flush() method and reorder to make mutable() always present; nice in theory but causes lost writes if flush() is retried

Revert "idea: inline the FlushHandle::flush() method and reorder to make mutable() always present; nice in theory but causes lost writes if flush() is retried"

Revert "WIP explore what happens if we make the fush loop infallible"

re-cherry-pick: Timeline::cancel sensitivity; this is sensitive but exposes again the mutable = None condition that we're fixing in this PR. We can go fix _that_ as welll, but, before we do that, e…

idea: remain read-available after Timeline::cancel has been cancelled; walreceiver shouldn't be retrying any writes, and will cause a panic if it does

WIP

WIP

it compiles

all rust tests pass

improve docs

linux