Issue 42040043: code review 42040043: go.tools/oracle: improvements to Emacs integration.
|
Created:
10 years, 4 months ago by adonovan Modified:
9 years, 4 months ago Reviewers:
Dominik Honnef CC:
Dominik Honnef, crawshaw1, golang-codereviews Visibility:
Public. |
Descriptiongo.tools/oracle: improvements to Emacs integration.
Overview:
- Eliminated go-oracle minor mode (another minor hurdle to adoption).
- Rebound some keys.
- Added var to toggle analysis of "reflect" package in PTA.
- "definition" command jumps directly to definition;
oracle buffer is not shown.
- Added "expand-selection" command: selects enclosing AST node.
- Factored out numerous utils:
go-oracle--{pos-flag,env,run,nop,jump-to,check-saved,json-query,query}
- Added menu.
The menu can be accessed via the menu-bar (if enabled) or via
a popup menu (C-down-mouse-3).
In menu-bar-mode, we additionally run a hook during redisplay
that causes menu items to be enabled/greyed out as appropriate
to the selection. It also rebuilds the expand-selection
submenu, which displays the enclosing AST and the current
package name.
Emacs doesn't provide the hook we really need to run some code
before the menu is displayed. We're forced to choose too
often (every redisplay, iff menu-bar-mode) or too infrequently
(not at all). Sigh.
Patch Set 1 #Patch Set 2 : diff -r 4e56dc65c681 https://code.google.com/p/go.tools #Patch Set 3 : diff -r 4e56dc65c681 https://code.google.com/p/go.tools #Patch Set 4 : diff -r e22a9223d6a9 https://code.google.com/p/go.tools #Patch Set 5 : diff -r e22a9223d6a9 https://code.google.com/p/go.tools #Patch Set 6 : diff -r e22a9223d6a9 https://code.google.com/p/go.tools #
Total comments: 23
Patch Set 7 : diff -r aca7c06737e1 https://code.google.com/p/go.tools #Patch Set 8 : diff -r aca7c06737e1 https://code.google.com/p/go.tools #MessagesTotal messages: 18
Hello dominik.honnef@gmail.com (cc: crawshaw@google.com, golang-dev@googlegroups.com), I'd like you to review this change to https://code.google.com/p/go.tools
Sign in to reply to this message.
LGTM with a few comments (and one actual issue) https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el File cmd/oracle/oracle.el (right): https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:70: ;; configuration. which is better than none. I like this idea https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:133: ;; TODO(adonovan): opt: this hook is fast but it runs *a lot*, Is there no way of running the hook only when the menu is actually opened? https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:138: ;; TODO(adonovan): rather than check the mode, only install the hook in go-mode. Yes please. I already hate the global gofmt-on-save hook we have that I only kept because of existing user configurations. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:148: ;; disturb the selected region. Will there ever be a way of sending the current buffer's content to the oracle (via stdin), so that it can work on modified buffers? https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:194: (goto-char (byte-to-position (1+ ,start))) holy nesting. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:334: (add-hook 'menu-bar-update-hook #'go-oracle--update-what) It took me quite a while to realize that this hook will not run when the menu bar at the top of Emacs is disabled, as it is in my setup. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:429: (define-key go-oracle-menu [select] It would be so cool if we could bind this to a key, which incrementally selects bigger and bigger parts of the ast related to the original selection. Or if we could open this specific menu as a text menu on a key press (like that of auto-completion), so you can select a menu point without having to use the mouse. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:439: ;; TODO(adonovan): is there a better keybinding for this menu? I hope so, C-S is quite awkward to press when ctrl is mapped to the capslock key. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:441: ;; cursor, if there no selection? That's what I always want for Position the cursor where?
Sign in to reply to this message.
https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el File cmd/oracle/oracle.el (right): https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:48: (define-key go-mode-map (kbd "C-c C-o t") #'go-oracle-describe) ; t for type Update: I'd like to change the prefix to "C-c o ...", to reduce the amount of keyboard-mashing slightly. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:70: ;; configuration. which is better than none. On 2013/12/14 04:29:07, Dominik Honnef wrote: > I like this idea Will leave as a TODO for now. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:133: ;; TODO(adonovan): opt: this hook is fast but it runs *a lot*, On 2013/12/14 04:29:07, Dominik Honnef wrote: > Is there no way of running the hook only when the menu is actually opened? I can't find a non-hackish way. The elisp docs (22.17.1.1 Simple Menu Items) say: The menu bar does not recalculate which items are enabled every time you look at a menu. This is because the X toolkit requires the whole tree of menus in advance. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:148: ;; disturb the selected region. On 2013/12/14 04:29:07, Dominik Honnef wrote: > Will there ever be a way of sending the current buffer's content to the oracle > (via stdin), so that it can work on modified buffers? Yes, it's on my list, but it will require a fair amount of work. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:334: (add-hook 'menu-bar-update-hook #'go-oracle--update-what) On 2013/12/14 04:29:07, Dominik Honnef wrote: > It took me quite a while to realize that this hook will not run when the menu > bar at the top of Emacs is disabled, as it is in my setup. Whoa, good catch... I too never use menu-bar-mode, but I enabled it for the purpose of developing this feature and so never noticed the problem. I'm not sure how best to proceed. Ideally menu-bar-mode should only do a 'what' query when actually needed, not on every redisplay, so the even when menu-bar-mode is enabled, the hook is slight overkill. Perhaps an idle timer can be made to work instead? When menu-bar-mode is disabled, the Go Oracle menu obtained from (C-S-mouse-down-1) still needs to be updated, but there's no hook at all here, and I haven't yet found something that can serve as one. I could use a Popup menu (29.15 Pop-Up Menus) but the docs advise: don't. Let me know if you have any ideas. I'll check with the local emacs gurus. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:429: (define-key go-oracle-menu [select] On 2013/12/14 04:29:07, Dominik Honnef wrote: > It would be so cool if we could bind this to a key, which incrementally selects > bigger and bigger parts of the ast related to the original selection. > > Or if we could open this specific menu as a text menu on a key press (like that > of auto-completion), so you can select a menu point without having to use the > mouse. I like this idea. I think it would be best to implement it as a separate command that makes its own "what" query each time it is invoked. This decouples it from the menu system entirely, which is simpler I think. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:439: ;; TODO(adonovan): is there a better keybinding for this menu? On 2013/12/14 04:29:07, Dominik Honnef wrote: > I hope so, C-S is quite awkward to press when ctrl is mapped to the capslock > key. It appears that C-S-down1 is already bound to a menu keymap that shows the mode-specific menus (e.g. Go oracle, Index, Buffers), so the two bindings seem to compete nondeterministically. I don't understand the exact mechanism (imenu, perhaps?) and this menu doesn't appear in the buffer's keymap. All of which is to say: we should definitely use a different binding. C-down3 opens the Go mode menu, of which the 'Go oracle' menu is a submenu. It's not as direct, but it's fewer things to remember, so perhaps its best not to add more bindings here. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:441: ;; cursor, if there no selection? That's what I always want for On 2013/12/14 04:29:07, Dominik Honnef wrote: > Position the cursor where? Under the current pointer location (as if preceded by a mouse-down-1 event). Perhaps your question was rhetorically suggesting that there is no hook that would allow us to run between the C-S-mouse-down-1 and the opening of the submenu. There doesn't seem to be.
Sign in to reply to this message.
On 2013/12/16 17:52:53, adonovan wrote: > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el > File cmd/oracle/oracle.el (right): > > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... > cmd/oracle/oracle.el:48: (define-key go-mode-map (kbd "C-c C-o t") > #'go-oracle-describe) ; t for type > Update: I'd like to change the prefix to "C-c o ...", to reduce the amount of > keyboard-mashing slightly. `C-c <single letter>` is reserved for binds that users may add in their configuration and shouldn't be used by modes/programs. http://www.gnu.org/software/emacs/manual/html_node/elisp/Key-Binding-Conventi... > > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... > cmd/oracle/oracle.el:334: (add-hook 'menu-bar-update-hook > #'go-oracle--update-what) > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > It took me quite a while to realize that this hook will not run when the menu > > bar at the top of Emacs is disabled, as it is in my setup. > > Whoa, good catch... I too never use menu-bar-mode, but I enabled it for the > purpose of developing this feature and so never noticed the problem. > > I'm not sure how best to proceed. Ideally menu-bar-mode should only do a 'what' > query when actually needed, not on every redisplay, so the even when > menu-bar-mode is enabled, the hook is slight overkill. Perhaps an idle timer > can be made to work instead? > > When menu-bar-mode is disabled, the Go Oracle menu obtained from > (C-S-mouse-down-1) still needs to be updated, but there's no hook at all here, > and I haven't yet found something that can serve as one. I could use a Popup > menu (29.15 Pop-Up Menus) but the docs advise: don't. > > Let me know if you have any ideas. I'll check with the local emacs gurus. No ideas here, sorry. An idle timer sounds very sub-optimal though. > > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... > cmd/oracle/oracle.el:429: (define-key go-oracle-menu [select] > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > It would be so cool if we could bind this to a key, which incrementally > selects > > bigger and bigger parts of the ast related to the original selection. > > > > Or if we could open this specific menu as a text menu on a key press (like > that > > of auto-completion), so you can select a menu point without having to use the > > mouse. > > I like this idea. I think it would be best to implement it as a separate > command that makes its own "what" query each time it is invoked. This decouples > it from the menu system entirely, which is simpler I think. Agreed. > > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... > cmd/oracle/oracle.el:439: ;; TODO(adonovan): is there a better keybinding for > this menu? > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > I hope so, C-S is quite awkward to press when ctrl is mapped to the capslock > > key. > > It appears that C-S-down1 is already bound to a menu keymap that shows the > mode-specific menus (e.g. Go oracle, Index, Buffers), so the two bindings seem > to compete nondeterministically. I don't understand the exact mechanism (imenu, > perhaps?) and this menu doesn't appear in the buffer's keymap. Interestingly, I don't seem to have anything bound to C-S-down1 until I load oracle.el. In retrospect I noticed the race as well, but thought I was being clumsy with the keys. > > All of which is to say: we should definitely use a different binding. > > C-down3 opens the Go mode menu, of which the 'Go oracle' menu is a submenu. > It's not as direct, but it's fewer things to remember, so perhaps its best not > to add more bindings here. That sounds acceptable to me, I think. I really never used menus in Emacs in my life, so I have no idea how other modes handle it. > > https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... > cmd/oracle/oracle.el:441: ;; cursor, if there no selection? That's what I > always want for > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > Position the cursor where? > > Under the current pointer location (as if preceded by a mouse-down-1 event). > > Perhaps your question was rhetorically suggesting that there is no hook that > would allow us to run between the C-S-mouse-down-1 and the opening of the > submenu. There doesn't seem to be. Any rhetoric was accidental. I do, however, disagree with the idea, in the same way that I taught middle click to yank where the point currently is, without moving it.
Sign in to reply to this message.
https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el File cmd/oracle/oracle.el (right): https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:48: (define-key go-mode-map (kbd "C-c C-o t") #'go-oracle-describe) ; t for type On 2013/12/16 17:52:54, adonovan wrote: > Update: I'd like to change the prefix to "C-c o ...", to reduce the amount of > keyboard-mashing slightly. Reverted to C-o per your advice. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:138: ;; TODO(adonovan): rather than check the mode, only install the hook in go-mode. On 2013/12/14 04:29:07, Dominik Honnef wrote: > Yes please. I already hate the global gofmt-on-save hook we have that I only > kept because of existing user configurations. I can't find any way to accomplish this. The hook is a global, not a buffer-local, so this check needs to be done in any case. Luckily it's cheap. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:334: (add-hook 'menu-bar-update-hook #'go-oracle--update-what) On 2013/12/16 17:52:54, adonovan wrote: > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > It took me quite a while to realize that this hook will not run when the menu > > bar at the top of Emacs is disabled, as it is in my setup. > > Whoa, good catch... I too never use menu-bar-mode, but I enabled it for the > purpose of developing this feature and so never noticed the problem. > > I'm not sure how best to proceed. Ideally menu-bar-mode should only do a 'what' > query when actually needed, not on every redisplay, so the even when > menu-bar-mode is enabled, the hook is slight overkill. Perhaps an idle timer > can be made to work instead? > > When menu-bar-mode is disabled, the Go Oracle menu obtained from > (C-S-mouse-down-1) still needs to be updated, but there's no hook at all here, > and I haven't yet found something that can serve as one. I could use a Popup > menu (29.15 Pop-Up Menus) but the docs advise: don't. > > Let me know if you have any ideas. I'll check with the local emacs gurus. No luck. I'm stuck with suboptimal behaviour for those users who use the popup menus but do not use `menu-bar-mode'. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:429: (define-key go-oracle-menu [select] On 2013/12/16 17:52:54, adonovan wrote: > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > It would be so cool if we could bind this to a key, which incrementally > selects > > bigger and bigger parts of the ast related to the original selection. > > > > Or if we could open this specific menu as a text menu on a key press (like > that > > of auto-completion), so you can select a menu point without having to use the > > mouse. > > I like this idea. I think it would be best to implement it as a separate > command that makes its own "what" query each time it is invoked. This decouples > it from the menu system entirely, which is simpler I think. Done. go-oracle-expand-selection, [C-c C-o e]. https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:439: ;; TODO(adonovan): is there a better keybinding for this menu? On 2013/12/16 17:52:54, adonovan wrote: > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > I hope so, C-S is quite awkward to press when ctrl is mapped to the capslock > > key. > > It appears that C-S-down1 is already bound to a menu keymap that shows the > mode-specific menus (e.g. Go oracle, Index, Buffers), so the two bindings seem > to compete nondeterministically. I don't understand the exact mechanism (imenu, > perhaps?) and this menu doesn't appear in the buffer's keymap. > > All of which is to say: we should definitely use a different binding. > > C-down3 opens the Go mode menu, of which the 'Go oracle' menu is a submenu. > It's not as direct, but it's fewer things to remember, so perhaps its best not > to add more bindings here. I removed this binding entirely since both C-down3 and C-S-down1 seem to work well enough. (They don't open the Oracle menu, but they open menus on which it is a submenu.) https://codereview.appspot.com/42040043/diff/90001/cmd/oracle/oracle.el#newco... cmd/oracle/oracle.el:441: ;; cursor, if there no selection? That's what I always want for On 2013/12/16 17:52:54, adonovan wrote: > On 2013/12/14 04:29:07, Dominik Honnef wrote: > > Position the cursor where? > > Under the current pointer location (as if preceded by a mouse-down-1 event). > > Perhaps your question was rhetorically suggesting that there is no hook that > would allow us to run between the C-S-mouse-down-1 and the opening of the > submenu. There doesn't seem to be. I've deleted this TODO as I think it's not feasible given Emacs' menu system.
Sign in to reply to this message.
Hello dominik.honnef@gmail.com (cc: crawshaw@google.com, golang-dev@googlegroups.com), Please take another look.
Sign in to reply to this message.
Two things: 1) for some reason, even with menu-bar-mode disabled, the context menu seems to update every time I press C-mouse3 and navigate to the Oracle submenu 2) I'm not sure if I broke my setup in the meantime, but using the Oracle only triggers a "peculiar error" message in the message buffer. It's not even a proper error, since it doesn't trigger the debugger. Invoking the oracle from the command line seems to work fine.
Sign in to reply to this message.
On 2013/12/17 19:36:24, Dominik Honnef wrote: > Two things: > > 1) for some reason, even with menu-bar-mode disabled, the context menu seems to > update every time I press C-mouse3 and navigate to the Oracle submenu Hmm, so it does. I had been working with the "Go oracle" menu previously bound to C-S-mouse-down1, which did not update, and never noticed that the mouse3 menu does seem to do a redisplay. It doesn't seem to use the usual keymap mechanism, and the code surrounding this seems to have changed a lot in recent versions of Emacs. (I'm using 24.3.1. What about you?) > 2) I'm not sure if I broke my setup in the meantime, but using the Oracle only > triggers a "peculiar error" message in the message buffer. It's not even a > proper error, since it doesn't trigger the debugger. Invoking the oracle from > the command line seems to work fine. Can you tell me exactly what you did?
Sign in to reply to this message.
On 2013/12/17 20:06:20, adonovan wrote: > On 2013/12/17 19:36:24, Dominik Honnef wrote: > > Two things: > > > > 1) for some reason, even with menu-bar-mode disabled, the context menu seems > to > > update every time I press C-mouse3 and navigate to the Oracle submenu > > Hmm, so it does. I had been working with the "Go oracle" menu previously bound > to C-S-mouse-down1, which did not update, and never noticed that the mouse3 menu > does seem to do a redisplay. It doesn't seem to use the usual keymap mechanism, > and the code surrounding this seems to have changed a lot in recent versions of > Emacs. (I'm using 24.3.1. What about you?) Same version here. > > > 2) I'm not sure if I broke my setup in the meantime, but using the Oracle only > > triggers a "peculiar error" message in the message buffer. It's not even a > > proper error, since it doesn't trigger the debugger. Invoking the oracle from > > the command line seems to work fine. > > Can you tell me exactly what you did? Just the usual workflow (loading oracle.el, opening a Go file, setting the scope, invoking C-c C-o d). I can see what the issue is now, though. The first time I invoked C-c C-o d, the oracle complained that it cannot find the package "dummyscope". Since I have debug-on-error set to t, that opened the debugger. The "peculiar error" stemmed from me forgetting to exit the debugger and invoking C-c C-o d again. What remains is the problem that the oracle complains about not finding a package called dummyscope.
Sign in to reply to this message.
On 17 December 2013 15:16, <dominik.honnef@gmail.com> wrote: > What > remains is the problem that the oracle complains about not finding a > package called dummyscope. > Have you built the oracle recently? Since a couple of days ago, query modes that don't need global analysis will ignore the args, guessing the main package from the -pos argument. However, main.go still checks that it's a nonempty slice. It's on my to-do list to fix (and document).
Sign in to reply to this message.
LGTM on the basis that this seems to be a bug in the oracle itself, not
oracle.el
On 2013/12/17 20:42:07, adonovan wrote:
> On 17 December 2013 15:16, <mailto:dominik.honnef@gmail.com> wrote:
>
> > What
> > remains is the problem that the oracle complains about not finding a
> > package called dummyscope.
> >
>
> Have you built the oracle recently? Since a couple of days ago, query
> modes that don't need global analysis will ignore the args, guessing the
> main package from the -pos argument. However, main.go still checks that
> it's a nonempty slice. It's on my to-do list to fix (and document).
dominikh@dominikh-pc!E:1 $ ~/go/bin/oracle -format=json
-pos=/home/dominikh/prj/go/src/honnef.co/go/id3/id3.go:\#512 definition
dummyscope
{
"mode": "definition",
"definition": {
"objpos": "/home/dominikh/go/src/pkg/log/log.go:281:6",
"desc": "func log.Println(v ...interface{})"
}
}%
dominikh@dominikh-pc $ readlink -f
/home/dominikh/prj/go/src/honnef.co/go/id3/id3.go
/storage/dominikh/projects/go/src/honnef.co/go/id3/id3.go
dominikh@dominikh-pc!E:1 $ ~/go/bin/oracle -format=json
-pos=/storage/dominikh/projects/go/src/honnef.co/go/id3/id3.go:\#512 definition
dummyscope
Error: cannot find package "dummyscope" in any of:
/home/dominikh/go/src/pkg/dummyscope (from $GOROOT)
/home/dominikh/prj/go/src/dummyscope (from $GOPATH).
My setup is that ~/prj is a symlink to ~/projects which is a symlink to
/storage/dominikh/projects. GOPATH is set to ~/prj/go – the Go tools seem to
handle this well; the oracle seems to be confused, but apparently only when the
definition mode is used? All other modes seem to work fine.
Sign in to reply to this message.
On 17 December 2013 15:59, <dominik.honnef@gmail.com> wrote: > LGTM on the basis that this seems to be a bug in the oracle itself, not > oracle.el > > On 2013/12/17 20:42:07, adonovan wrote: > > On 17 December 2013 15:16, <mailto:dominik.honnef@gmail.com> wrote: >> > > > What >> > remains is the problem that the oracle complains about not finding a >> > package called dummyscope. >> > >> > > Have you built the oracle recently? Since a couple of days ago, query >> modes that don't need global analysis will ignore the args, guessing >> > the > >> main package from the -pos argument. However, main.go still checks >> > that > >> it's a nonempty slice. It's on my to-do list to fix (and document). >> > > dominikh@dominikh-pc!E:1 $ ~/go/bin/oracle -format=json > -pos=/home/dominikh/prj/go/src/honnef.co/go/id3/id3.go:\#512 definition > dummyscope > { > "mode": "definition", > "definition": { > "objpos": "/home/dominikh/go/src/pkg/log/log.go:281:6", > "desc": "func log.Println(v ...interface{})" > } > }% > > dominikh@dominikh-pc $ readlink -f > /home/dominikh/prj/go/src/honnef.co/go/id3/id3.go > /storage/dominikh/projects/go/src/honnef.co/go/id3/id3.go > dominikh@dominikh-pc!E:1 $ ~/go/bin/oracle -format=json > -pos=/storage/dominikh/projects/go/src/honnef.co/go/id3/id3.go:\#512 > definition dummyscope > Error: cannot find package "dummyscope" in any of: > /home/dominikh/go/src/pkg/dummyscope (from $GOROOT) > /home/dominikh/prj/go/src/dummyscope (from $GOPATH). > > > My setup is that ~/prj is a symlink to ~/projects which is a symlink to > /storage/dominikh/projects. GOPATH is set to ~/prj/go – the Go tools > seem to handle this well; the oracle seems to be confused, but > apparently only when the definition mode is used? All other modes seem > > to work fine. > > https://codereview.appspot.com/42040043/ > Ah, thanks for digging deeper. The heuristic the oracle uses to ascertain the package from the -pos flag is incomplete, and so it's falling back to using the pointer analysis scope, which doesn't exist. I should improve the heuristic and also eliminate the dummyscope hack. I'm taking vacation till Dec 30; I'll fix this on my return. happy holidays alan
Sign in to reply to this message.
Replacing golang-dev with golang-codereviews.
Sign in to reply to this message.
Replacing golang-dev with golang-codereviews.
Sign in to reply to this message.
Replacing golang-dev with golang-codereviews.
Sign in to reply to this message.
What happened to this?
Sign in to reply to this message.
It got put on hold because I discovered some bugs in the way Emacs menus work. I will dig it up again; not this week though. On 13 June 2014 05:10, <dominik.honnef@gmail.com> wrote: > What happened to this? > > https://codereview.appspot.com/42040043/ >
Sign in to reply to this message.
R=close To the author of this CL: The Go project has moved to Gerrit Code Review. If this CL should be continued, please see the latest version of https://golang.org/doc/contribute.html for instructions on how to set up Git and the Go project's Gerrit codereview plugin, and then create a new change with your current code. If there has been discussion on this CL, please give a link to it (golang.org/cl/42040043 is best) in the description in your new CL. Thanks very much.
Sign in to reply to this message.
|
