[PATCH 4 of 4] largefiles: makes verify batching stat calls to remote

[Yuya Nishihara]

On Mon, 02 May 2016 22:19:12 +0200, liscju wrote:
> # HG changeset patch
> # User liscju <piotr.listkiewicz at gmail.com>
> # Date 1462210631 -7200
> #      Mon May 02 19:37:11 2016 +0200
> # Branch stable
> # Node ID 5ca4f289ae02c580e19eafe7bf6168bda92664eb
> # Parent  889c3537ab941231ebc6881af6c1be6f0321294b
> largefiles: makes verify batching stat calls to remote
> 
> Instead of sending stat calls for each file separately, it send
> one batch call with stat invocations for all files.
> 
> diff -r 889c3537ab94 -r 5ca4f289ae02 hgext/largefiles/remotestore.py
> --- a/hgext/largefiles/remotestore.py	Mon May 02 18:52:49 2016 +0200
> +++ b/hgext/largefiles/remotestore.py	Mon May 02 19:37:11 2016 +0200
> @@ -67,28 +67,31 @@ class remotestore(basestore.basestore):
>  
>      def _verify(self, revs, contents, verified):
>          failed = False
> +        # filestocheck is list of (cset, filename, expectedhash)
> +        filestocheck = []
>          for rev in revs:
>              cctx = self.repo[rev]
>              cset = "%d:%s" % (cctx.rev(), node.short(cctx.node()))
>  
>              for standin in cctx:
> -                if self._verifyfile(cctx, cset, contents, standin, verified):
> -                    failed = True
> +                filename = lfutil.splitstandin(standin)
> +                if filename:
> +                    fctx = cctx[standin]
> +                    key = (filename, fctx.filenode())
> +                    expectedhash = fctx.data()[0:40]
> +                    verified.add(key)
> +                    filestocheck.append((cset, filename, expectedhash))

filestocheck would have duplicated entries if you do "verify --large --lfa".

> +        expectedhashes = [expecthash for _, _, expecthash in filestocheck]

Hiding i18n._() isn't good.

> -    def _verifyfile(self, cctx, cset, contents, standin, verified):
> -        filename = lfutil.splitstandin(standin)
> -        if not filename:
> -            return False
> -        fctx = cctx[standin]
> -        key = (filename, fctx.filenode())
> -        if key in verified:
> -            return False

Before, duplicated entries would be dropped here.

I have no idea if the direction of this series is good. Since 'filestocheck'
can be built locally, another option would be to make _verifyfiles(filestocheck)
an abstract interface.