code review 3522041: govet: a new static checker for Go programs.

govet: a new static checker for Go programs.
At the moment, and for the forseeable future, it only checks arguments to print calls.

[r, 2010-12-08 21:22:34]

Hello rsc, gri (cc: golang-dev@googlegroups.com),

I'd like you to review this change.

[r, 2010-12-08 21:26:25]

Hello rsc, gri (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc1, 2010-12-08 21:33:28]

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/doc.go
File src/cmd/gocheck/doc.go (right):

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/doc.go#newcode16
src/cmd/gocheck/doc.go:16: a format descriptor string in the manner of
fmt.Printf.
a
If not, gocheck warns about arguments that look like format descriptor strings.
.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:27: var helpString = `By default gocheck checks for
functions from
This is already out of date (-funcs below).
Leave for [godoc gocheck] to print instead?

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:48: if *help {
I'm uncomfortable with this precedent (see above).

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:78: die(`illegal format for "Func:N" argument "%s";
%s`, name, err)
s/"%s"/%q/

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:93: for i := 0; i < flag.NArg(); i++ {
for _, arg := range flag.Args() {

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:116: fmt.Fprintf(os.Stderr, "gocheckprint:
"+format+"\n", args...)
s2/print//

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:120: // Println is fmt.Println guarded by --v.
s/--/-/
(not incorrect but unnecessary)

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:128: // Printf is fmt.Printf guarded by --v.
s/--/-/

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:161: func (f *File) checkFile(name string, file
*ast.File) {
I think much of this can go away if you use ast.Walk.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:340: "Printf": 0, "Println": 0,
s/f//
.+1s/f//

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:361: arg := call.Args[skip]
check len(call.Args)

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:373: f.Badf(call.Pos(), "no formatting directive in
%s call", name)
I don't think this is bad.
Programs that, say, use Printf consistently seem fine.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:397: var percent = []byte{'%'}
can use IndexByte below instead

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:405: args := call.Args
check len(call.Args)

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:408: f.Badf(call.Pos(), "no args in %s call", name)
I don't think this is worth complaining about
until the code can distinction fmt.Print() from log.Print()

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:415: f.Badf(call.Pos(), "formatting directive in %s
call", name)
s/formatting/possible &

[r, 2010-12-08 21:46:25]

Hello rsc, gri (cc: golang-dev@googlegroups.com),

Please take another look.

[r, 2010-12-08 21:46:38]

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:27: var helpString = `By default gocheck checks for
functions from
On 2010/12/08 21:33:28, rsc1 wrote:
> This is already out of date (-funcs below).
> Leave for [godoc gocheck] to print instead?

i'm torn too. i like the command to give me information. plus people don't seem
to know that godoc describes commands.  but ok.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:78: die(`illegal format for "Func:N" argument "%s";
%s`, name, err)
On 2010/12/08 21:33:28, rsc1 wrote:
> s/"%s"/%q/

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:93: for i := 0; i < flag.NArg(); i++ {
On 2010/12/08 21:33:28, rsc1 wrote:
> for _, arg := range flag.Args() {

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:116: fmt.Fprintf(os.Stderr, "gocheckprint:
"+format+"\n", args...)
On 2010/12/08 21:33:28, rsc1 wrote:
> s2/print//

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:120: // Println is fmt.Println guarded by --v.
On 2010/12/08 21:33:28, rsc1 wrote:
> s/--/-/
> (not incorrect but unnecessary)

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:128: // Printf is fmt.Printf guarded by --v.
On 2010/12/08 21:33:28, rsc1 wrote:
> s/--/-/

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:161: func (f *File) checkFile(name string, file
*ast.File) {
On 2010/12/08 21:33:28, rsc1 wrote:
> I think much of this can go away if you use ast.Walk.

doesn't really seem much of an improvement to go that way. i still need to write
lots of fiddly bits.  i like it as is.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:340: "Printf": 0, "Println": 0,
On 2010/12/08 21:33:28, rsc1 wrote:
> s/f//
> .+1s/f//

oops. i bet another CL is coming

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:361: arg := call.Args[skip]
On 2010/12/08 21:33:28, rsc1 wrote:
> check len(call.Args)

check it how?

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:373: f.Badf(call.Pos(), "no formatting directive in
%s call", name)
On 2010/12/08 21:33:28, rsc1 wrote:
> I don't think this is bad.
> Programs that, say, use Printf consistently seem fine.
> 

it's bad if you have more than one argument, which is what this checking.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:397: var percent = []byte{'%'}
On 2010/12/08 21:33:28, rsc1 wrote:
> can use IndexByte below instead

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:405: args := call.Args
On 2010/12/08 21:33:28, rsc1 wrote:
> check len(call.Args)

check it how?

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:408: f.Badf(call.Pos(), "no args in %s call", name)
On 2010/12/08 21:33:28, rsc1 wrote:
> I don't think this is worth complaining about
> until the code can distinction fmt.Print() from log.Print()

Done.

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:415: f.Badf(call.Pos(), "formatting directive in %s
call", name)
On 2010/12/08 21:33:28, rsc1 wrote:
> s/formatting/possible &

Done.

[r, 2010-12-08 21:46:53]

Hello rsc, gri (cc: golang-dev@googlegroups.com),

Please take another look.

[rsc1, 2010-12-08 21:59:17]

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:361: arg := call.Args[skip]
On 2010/12/08 21:46:38, r wrote:
> On 2010/12/08 21:33:28, rsc1 wrote:
> > check len(call.Args)
> 
> check it how?

if len(call.Args) <= skip {

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:405: args := call.Args
On 2010/12/08 21:46:38, r wrote:
> On 2010/12/08 21:33:28, rsc1 wrote:
> > check len(call.Args)
> 
> check it how?

Never mind on this one.

[rsc1, 2010-12-08 22:04:06]

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/1/src/cmd/gocheck/gocheck.go#newco...
src/cmd/gocheck/gocheck.go:161: func (f *File) checkFile(name string, file
*ast.File) {
On 2010/12/08 21:46:38, r wrote:
> On 2010/12/08 21:33:28, rsc1 wrote:
> > I think much of this can go away if you use ast.Walk.
> 
> doesn't really seem much of an improvement to go that way. i still need to
write
> lots of fiddly bits.  i like it as is.

you can pass an *ast.File to ast.Walk and then in
the walker look for *ast.CallExpr and call your checkCallExpr.
if you did that, 

checkFile
checkDecl
checkFuncDecl
checkBlock
checkStmtList
checkStmt
checkExprList
checkExpr

would all just disappear.
that's 150 lines that won't need maintenance
if something other than the form of a call
expression changes.  it seems like a big deal.

[gri, 2010-12-08 22:13:39]

Some comments. Will look at it again more precisely later.

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:28: func Usage() {
s/Usage/usage/ ?

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:37: fs *token.FileSet
if you are only operating on a single file, use

file *token.File

instead. It also supports a Position function and it's more efficient

(if you think it may change to multiple files, keep the fileset - either way
it's easy to change)

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:79: fs := token.NewFileSet()
I'd use a global fset. If you call doFile multiple times there's no need to
re-allocate a new fset every single time.

In fact, if you use a global fset, you may not need the per file data structure.

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:80: fs.AddFile(name, fs.Base(), 1e6) // TODO: how big
should the size argument be?
this is not needed - the parser adds the file to the file set (maybe it
shouldn't, but that's how it is now)

it will read the entire file into memory and then know the size for AddFile

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:85: file := &File{fs}
You may not need this. If you want to keep it (no global fset); or if you want
to keep track of the per-file *token.File (a bit faster Position function than
the one on fset), then write:

file := &File{fset.File(parsedFile.Pos())}

(This will take a legal position in the parsedFile - the position of the packake
keyword - and get the corresponding file back. Since the parser does the AddFile
internally, and it doesn't return it explicitly, not store it in the AST, this
is a simple mechanism to get it back).

http://codereview.appspot.com/3522041/diff/15001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:173: func (f *File) checkStmt(stmt ast.Stmt) {
Instead of writing all this tree traversal yourself, why not use the ast.Walk
function and just implement the few cases you actually need?

It will take care of the right nil checks etc.

(Note: I am in the middle of fixing a bug in ast.Walk, but the interface should
remain the same).

This may cut away a lot of code.

There's also ast.Inspect with a slightly simpler interface.

[r, 2010-12-08 22:34:26]

Hello rsc, gri (cc: golang-dev@googlegroups.com),

Please take another look.

[r, 2010-12-08 22:34:46]

I did the rewrite. It's cleaner code; it's also 50% slower, which is
interesting.  Gri??

-rob

[niemeyer, 2010-12-08 22:37:26]

Hi Rob,

If possible, I'd appreciate if another name was selected for the tool.

gocheck is being developed for several months, and was already announced and
commented upon in the mailing list:

http://labix.org/gocheck
http://blog.labix.org/2010/11/06/gocheck-a-rich-testing-library-for-go
http://launchpad.net/gocheck

[gri, 2010-12-08 22:38:28]

How about golint?
- gri

On Wed, Dec 8, 2010 at 2:37 PM, <n13m3y3r@gmail.com> wrote:

> Hi Rob,
>
> If possible, I'd appreciate if another name was selected for the tool.
>
> gocheck is being developed for several months, and was already announced
> and commented upon in the mailing list:
>
> http://labix.org/gocheck
> http://blog.labix.org/2010/11/06/gocheck-a-rich-testing-library-for-go
> http://launchpad.net/gocheck
>
>
> http://codereview.appspot.com/3522041/
>

[r, 2010-12-08 22:43:10]

lint was a total piece of crap with horrible properties that corrupted
programmers for a generation. golint is an obvious name but i can't
bear to associate this program with that one.  moreover, lint did
vastly more than gocheck, and more than gocheck should or will since
we have types and a decent compiler, so i don't want people to think
about them together.

finally, the joke in lint's name always bothered me.

otherwise, a fine suggestion.

-rob

[niemeyer, 2010-12-08 22:49:21]

> otherwise, a fine suggestion.

golint
goverify
gotest --check
goguard
gostatic
gowell
gopure
gosafe
gofit

If you need more, please let me know.  Let's just avoid
the conflict, please.

[iant2, 2010-12-08 22:52:25]

n13m3y3r@gmail.com writes:

>> otherwise, a fine suggestion.
>
> golint
> goverify
> gotest --check
> goguard
> gostatic
> gowell
> gopure
> gosafe
> gofit

Or make it an option for gofmt.

Ian

[r, 2010-12-08 22:59:22]

On Wed, Dec 8, 2010 at 5:52 PM, Ian Lance Taylor <iant@google.com> wrote:
> n13m3y3r@gmail.com writes:
>
>>> otherwise, a fine suggestion.
>>
>> golint
>> goverify
>> gotest --check
>> goguard
>> gostatic
>> gowell
>> gopure
>> gosafe
>> gofit
>
> Or make it an option for gofmt.
>
> Ian
>

gofmt is already pretty heavily burdened. i was thinking this could
take some of the load off. it's a tiny program that can expand as
needed.

but it does make sense to fold it in.

-rob

[r, 2010-12-08 23:00:28]

oh that thing. dammit.

-rob

[r, 2010-12-08 23:00:50]

Hello rsc, gri, niemeyer, iant2 (cc: golang-dev@googlegroups.com),

Please take another look.

[r, 2010-12-08 23:01:19]

PTAL while we consider throwing it all away.

-rob

[r, 2010-12-08 23:11:07]

on further thought, gofmt doesn't seem like a good match.  its job is
not to warn about problems but to reformat the code. what would gofmt
do if it found an error? and it can't be automatic because this
program doesn't (and can't) do much semantic or dynamic stuff; it's
all simple static pattern matches.

i believe it makes sense to have a separate tool with potentially
expensive yet inaccurate checks for sanity.

i just need a name.

-rob

[gri, 2010-12-08 23:11:21]

http://codereview.appspot.com/3522041/diff/33001/src/cmd/gocheck/gocheck.go
File src/cmd/gocheck/gocheck.go (right):

http://codereview.appspot.com/3522041/diff/33001/src/cmd/gocheck/gocheck.go#n...
src/cmd/gocheck/gocheck.go:147: return f
I think you ought to be able to speed this up:

Whenever you have a node where you know that it cannot _contain_ a call (all
decls except for function decls) you can return nil as result. This will prevent
that subtree from being visited. As it is, the entire tree is traversed even
where it makes no sense. This may explain the slowdown to some extent.

[gri, 2010-12-08 23:12:56]

agreed - I don't want gofmt to become the kitchen sink for everything source
code related

gosane?

- gri

On Wed, Dec 8, 2010 at 3:11 PM, Rob 'Commander' Pike <r@golang.org> wrote:

> on further thought, gofmt doesn't seem like a good match.  its job is
> not to warn about problems but to reformat the code. what would gofmt
> do if it found an error? and it can't be automatic because this
> program doesn't (and can't) do much semantic or dynamic stuff; it's
> all simple static pattern matches.
>
> i believe it makes sense to have a separate tool with potentially
> expensive yet inaccurate checks for sanity.
>
> i just need a name.
>
> -rob
>

[rog, 2010-12-08 23:14:39]

what about "gowarn"?

i.e. a tool that gives you the warnings that the
compiler does not.

On 8 December 2010 23:11, Rob 'Commander' Pike <r@golang.org> wrote:
> on further thought, gofmt doesn't seem like a good match.  its job is
> not to warn about problems but to reformat the code. what would gofmt
> do if it found an error? and it can't be automatic because this
> program doesn't (and can't) do much semantic or dynamic stuff; it's
> all simple static pattern matches.
>
> i believe it makes sense to have a separate tool with potentially
> expensive yet inaccurate checks for sanity.
>
> i just need a name.
>
> -rob
>

[r, 2010-12-08 23:15:39]

Hello rsc, gri, niemeyer, iant2, rog (cc: golang-dev@googlegroups.com),

Please take another look.

[r, 2010-12-08 23:16:09]

Hello rsc, gri, niemeyer, iant2, rog (cc: golang-dev@googlegroups.com),

Please take another look.

[lstoakes, 2010-12-08 23:16:56]

gocop? .net has 'fxcop' which performs static analysis, seems like a
fair analogy!


On 8 December 2010 23:12, Robert Griesemer <gri@golang.org> wrote:
> agreed - I don't want gofmt to become the kitchen sink for everything source
> code related
> gosane?
> - gri
>
> On Wed, Dec 8, 2010 at 3:11 PM, Rob 'Commander' Pike <r@golang.org> wrote:
>>
>> on further thought, gofmt doesn't seem like a good match.  its job is
>> not to warn about problems but to reformat the code. what would gofmt
>> do if it found an error? and it can't be automatic because this
>> program doesn't (and can't) do much semantic or dynamic stuff; it's
>> all simple static pattern matches.
>>
>> i believe it makes sense to have a separate tool with potentially
>> expensive yet inaccurate checks for sanity.
>>
>> i just need a name.
>>
>> -rob
>
>



-- 
Lorenzo Stoakes
http://www.codegrunt.co.uk

[r, 2010-12-08 23:17:22]

i chose 'govet', since it vets the code.

that is the end of the naming discussion.

-rob

[niemeyer, 2010-12-08 23:26:47]

> i chose 'govet', since it vets the code.
> 
> that is the end of the naming discussion.

Phew.  Thanks!

I'm glad we can all go to the vet now. Oh, wait.. crap.

[gri, 2010-12-08 23:53:19]

LGTM

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/doc.go
File src/cmd/govet/doc.go (right):

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/doc.go#newcode29
src/cmd/govet/doc.go:29: is zero-indexed argument position of the first argument
s/is/is the/

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/doc.go#newcode29
src/cmd/govet/doc.go:29: is zero-indexed argument position of the first argument
also should it be "zero-based" ? Not sure what zero-indexed means.

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/doc.go#newcode31
src/cmd/govet/doc.go:31: arguemnt for non-formatted prints.
s/arguemnt/argument/

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/doc.go#newcode31
src/cmd/govet/doc.go:31: arguemnt for non-formatted prints.
perhaps an example using fmt.Printf?

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/govet.go
File src/cmd/govet/govet.go (right):

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/govet.go#newco...
src/cmd/govet/govet.go:72: }
It would be nice if it could operate on directories like gofmt. Maybe another
CL.

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/govet.go#newco...
src/cmd/govet/govet.go:145: f.checkCallExpr(n)
add a TODO: should return nil for nodes that cannot contain a CallExpr - will
shortcut traversal

http://codereview.appspot.com/3522041/diff/46001/src/cmd/govet/govet.go#newco...
src/cmd/govet/govet.go:237: var terminalNewline = []byte{'\\', 'n', '"'} // \n
at end of interpreted string
[]byte(`\n"`) ?

[jacekm, 2010-12-09 00:54:32]

http://codereview.appspot.com/3522041/diff/57001/src/cmd/govet/govet.go
File src/cmd/govet/govet.go (right):

http://codereview.appspot.com/3522041/diff/57001/src/cmd/govet/govet.go#newco...
src/cmd/govet/govet.go:27: // If --help is set, it prints a longer description.
Delete because --help is gone.

[cw, 2010-12-09 06:49:55]

http://codereview.appspot.com/3522041/diff/63001/src/cmd/make.bash
File src/cmd/make.bash (right):

http://codereview.appspot.com/3522041/diff/63001/src/cmd/make.bash#newcode24
src/cmd/make.bash:24: for i in cc ${O}l ${O}a ${O}c gc ${O}g cov godefs gopack
gotest govet nm prof
this needs to be removed and rather have something like:

diff -r a5eebdd09522 src/pkg/Makefile
--- a/src/pkg/Makefile  Wed Dec 08 21:36:56 2010 -0800
+++ b/src/pkg/Makefile  Wed Dec 08 22:49:14 2010 -0800
@@ -139,6 +139,7 @@
        ../cmd/goinstall\
        ../cmd/goyacc\
        ../cmd/hgpatch\
+       ../cmd/govet\
 
 NOTEST=\
        debug/proc\

[r, 2010-12-09 16:17:17]

Hello rsc, gri, niemeyer, iant2, rog, lstoakes, jacek.masiulaniec@gmail.com, cw
(cc: golang-dev@googlegroups.com),

Please take another look.

[rsc1, 2010-12-09 17:14:49]

LGTM

[r, 2010-12-09 17:37:25]

*** Submitted as http://code.google.com/p/go/source/detail?r=752e82357e09 ***

govet: a new static checker for Go programs.
At the moment, and for the forseeable future, it only checks arguments to print
calls.

R=rsc, gri, niemeyer, iant2, rog, lstoakes, jacek.masiulaniec, cw
CC=golang-dev
http://codereview.appspot.com/3522041