From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <e@80x24.org>
X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on dcvr.yhbt.net
X-Spam-Level: 
X-Spam-ASN:  
X-Spam-Status: No, score=-4.0 required=3.0 tests=ALL_TRUSTED,BAYES_00
	shortcircuit=no autolearn=ham autolearn_force=no version=3.4.2
Received: from localhost (dcvr.yhbt.net [127.0.0.1])
	by dcvr.yhbt.net (Postfix) with ESMTP id 671581F463;
	Thu, 26 Sep 2019 03:03:57 +0000 (UTC)
Date: Thu, 26 Sep 2019 03:03:57 +0000
From: Eric Wong <e@80x24.org>
To: edef <edef@edef.eu>
Cc: meta@public-inbox.org, hi@alyssa.is
Subject: Re: [PATCH 0/1] Fix broken clone URLs due to SCRIPT_NAME getting
 reset
Message-ID: <20190926030357.GA21009@dcvr>
References: <cover.0f97c47bb88db8b875be7497289d8fedd3b11991.1569296942.git-series.edef@edef.eu>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <cover.0f97c47bb88db8b875be7497289d8fedd3b11991.1569296942.git-series.edef@edef.eu>
List-Id: <meta.public-inbox.org>

edef <edef@edef.eu> wrote:
> We're trying to get public-inbox working with a PSGI file that mounts
> it to a subdirectory. This seems like it's intended to be a supported
> use case, with stuff paying attention to SCRIPT_NAME and all when
> generating URLs.
> 
> However, Plack::App::URLMap seems determined to reset SCRIPT_NAME
> before getline gets called:
> 
>     my $orig_path_info   = $env->{PATH_INFO};
>     my $orig_script_name = $env->{SCRIPT_NAME};
> 
>     $env->{PATH_INFO}  = $path;
>     $env->{SCRIPT_NAME} = $script_name . $location;
>     return $self->response_cb($app->($env), sub {
>         $env->{PATH_INFO} = $orig_path_info;
>         $env->{SCRIPT_NAME} = $orig_script_name;
>     });

Sounds like a familiar problem to me :x

> I'm not sure whether public-inbox or Plack is in the wrong here, but
> the timing works out poorly. By the time
> PublicInbox::WwwStream::_html_end gets invoked SCRIPT_NAME is blank,
> and the wrong URLs get generated.
> 
> Copying env seems to fix it, and that's what the attached patch does.
> I'm pretty sure this is the wrong approach, but it seems to work.

Yeah, it's a big hash and not needed to copy the whole thing.

I gotta run, now, but I think the patch below will work for you
by precalculating base_url up front.   Can you confirm?  Thanks.

Also, I suspect the mbox Archived-At headers could be wrong
and need a similar change...  Maybe Atom feeds, too.

diff --git a/lib/PublicInbox/WwwStream.pm b/lib/PublicInbox/WwwStream.pm
index e0823c8d..b240c071 100644
--- a/lib/PublicInbox/WwwStream.pm
+++ b/lib/PublicInbox/WwwStream.pm
@@ -19,7 +19,17 @@ sub close {}
 
 sub new {
 	my ($class, $ctx, $cb) = @_;
-	bless { nr => 0, cb => $cb || *close, ctx => $ctx }, $class;
+
+	my $env = $ctx->{env};
+	my $ibx = $ctx->{-inbox};
+	my $base_url = $ibx->base_url($env);
+	chop $base_url; # no trailing slash for clone
+	bless {
+		nr => 0,
+		cb => $cb || *close,
+		ctx => $ctx,
+		base_url => $base_url,
+	}, $class;
 }
 
 sub response {
@@ -83,8 +93,7 @@ sub _html_end {
 	my $desc = ascii_html($ibx->description);
 
 	my (%seen, @urls);
-	my $http = $ibx->base_url($ctx->{env});
-	chop $http; # no trailing slash for clone
+	my $http = $self->{base_url};
 	my $max = $ibx->max_git_epoch;
 	my $dir = (split(m!/!, $http))[-1];
 	if (defined($max)) { # v2