Re: [privoxy-devel] TODO list proposal

[Fabian K. <fk...@fa...>]

Lee <le...@gm...> wrote:

> On 4/25/09, Fabian Keil <fk...@fa...> wrote:
> 
> >> The test nicely highlights that the current template
> >> system is ridiculously slow, though (item 50).
> >
> > On a second thought it actually doesn't. The bottleneck
> > here is probably Privoxy-Regression-Test itself, which has
> > to parse Privoxy's response.

And if you assume ...

I just ran Privoxy-Regression-Test through SmallProf and it
seems to spend most of the time waiting for the CGI page:

fk@TP51 ~/test/privoxy $sort -k 2nr,2 smallprof.out | head
      516  82.23432  12.85156  1212:        @buffer = `$curl_line`;
    42532   0.22723   0.71875   773:        $final_results_reached = 1 if
    42532   0.17074   0.67969   775:        next unless ($final_results_reached);
      476   0.12992   0.07031  1038:    my @buffer = @{$buffer_ref};
     9377   0.04342   0.15625   782:            if (defined $parameter) {
        1   0.03176   0.00000  1104:    l(LL_ERROR, "No curl found.") unless
     4996   0.02806   0.04688   222:    s@&quot;@"@g;
      476   0.02505   0.01562  1042:        if (/^HTTP\/\d\.\d (\d{3})/) {
    10049   0.02505   0.21875   778:        if (m@<br>([-+])<a.*>([^>]*)</a>(?:
     9377   0.02440   0.16406   780:            my $parameter = $3;

However I also compared fetching a CGI page from Privoxy with
fetching the result from a local HTTP server, and the difference
isn't that big.

It looks like speeding up the template system may be useful when
running Privoxy in valgrind, but less useful from a user's point
of view.

> Would it be helpful to add another p.p URL that returns only enough
> info for Privoxy-Regression-Test to do it's job?  Seems like all it
> would need is the "final results" section without all the hyperlinks

According to SmallProf it probably wouldn't.

Fabian