x/tools/go/pointer: isolate, tag, and delete the pointer analysis #59676
Comments
|
All of the maintenance issues given in the proposal are somewhat to be expected with whole program static analysis. These are fixable with an additional investment of time. It is however clear that maintaining the package has not been a high priority for a while and issues with the package are not getting addressed. I think accepting this proposal would set better expectations for users going forward. |
|
This proposal has been added to the active column of the proposals project |
|
Based on the discussion above, this proposal seems like a likely accept. |
|
No change in consensus, so accepted. 🎉 |
rsc
changed the title
|
Change https://go.dev/cl/499695 mentions this issue: |
|
Change https://go.dev/cl/499696 mentions this issue: |
This change removes the -algo=pta option from cmd/callgraph, and all the subcommands of cmd/guru, that use pointer analysis. These features have been poorly supported for a long time, and the pointer analysis package is about to be tagged and deleted. Updates golang/go#59676 Change-Id: Id4ded651b8385c588991d01377b2f087d14ae191 Reviewed-on: https://go-review.googlesource.com/c/tools/+/499696 gopls-CI: kokoro <noreply+kokoro@google.com> Reviewed-by: Bryan Mills <bcmills@google.com> Run-TryBot: Alan Donovan <adonovan@google.com> TryBot-Result: Gopher Robot <gobot@golang.org>
The pointer analysis is deprecated. Once this module is tagged, we will delete it. Clients will be able to use only the last version. Updates golang/go#59676 Change-Id: If6e2214d964f0bbd1267634a1ad5c0d3166ecf0e Reviewed-on: https://go-review.googlesource.com/c/tools/+/499695 Run-TryBot: Alan Donovan <adonovan@google.com> Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com> Reviewed-by: Robert Findley <rfindley@google.com> TryBot-Result: Gopher Robot <gobot@golang.org> Auto-Submit: Alan Donovan <adonovan@google.com>
|
Change https://go.dev/cl/499756 mentions this issue: |
|
Change https://go.dev/cl/499757 mentions this issue: |
This makes it so that 'go install golang.org/x/tools/go/pointer@latest'
will continue to work. ('go install pkg@version' does not allow
replace directives.)
Updates golang/go#59676
Change-Id: I7dcb95a5730bd68d42a5873f9f9cec6bef773205
Reviewed-on: https://go-review.googlesource.com/c/tools/+/499756
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Alan Donovan <adonovan@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
golang.org/x/tools/go/pointer is a pointer analysis for Go. it was originally developed for cmd/guru, which itself is moribund, succeeded by the much more sophisticated gopls. Pointer analysis in an IDE was an interesting idea, but it proved much less useful than expected: it is slow and scales poorly, and even when correct its results were often hard to comprehend. It requires some notion of a main (or test) entry point, which means it requires more configuration than most users are willing to spend on it. It is unsound: like most static analyses, it can't track information flow and call paths through reflection, even though these are widely used by text/template, encoding/json, encoding/gob, and many other packages. And it is inherently fragile, since it requires annotations for non-Go functions such as C and assembly. Few users are willing, able, or aware of the need to write such annotations, and the few annotations that do exist, for functions in the runtime and standard library, become stale almost as soon as they are updated.
We propose to create a new module for golang.org/x/tools/go/pointer, tag it, then delete it. At that point, applications that wish to depend on the last tagged version (after updating their go.mod file) may continue to do so, but there will be no more fixes to the package. Perhaps someone in the Go community will fork it and give it a new life.
It is currently imported by 233 projects.
See also: #59056
The text was updated successfully, but these errors were encountered: