Re: [pcre-dev] 3x-4x slowdown in pcre_match

Top Page

Reply to this message
Author: enh
Date:  
To: Petr Pisar, pcre-dev
Subject: Re: [pcre-dev] 3x-4x slowdown in pcre_match
On Wed, Apr 22, 2020 at 12:49 AM Petr Pisar via Pcre-dev
<pcre-dev@???> wrote:
>
> On Tue, Apr 21, 2020 at 04:21:44PM -0700, enh via Pcre-dev wrote:
> > On Tue, Apr 21, 2020 at 12:53 AM <ph10@???> wrote:
> > >
> > > On Mon, 20 Apr 2020, enh via Pcre-dev wrote:
> > >
> > > > > Thank you. I will deal with this in a day or two (diverted elsewhere at
> > > > > the moment) along with several other minor tweaks that have just
> > > > > arrived.
> > > >
> > > > thanks! (i was worried that the patch got mangled by the mailing list
> > > > because it looks a mess on
> > > > https://lists.exim.org/lurker/message/20200416.220034.04568e78.en.html
> > > > but i'll take this as a sign that it got through okay! let me know if
> > > > i should send it some other way. gmail's not the best for patches...)
> > >
> > > Oh, I hadn't actually looked at it, but now I have saved the file and it
> > > looks OK. Oddly, it looks reasonable to me on that mailing list URL as
> > > well (using Firefox on Arch Linux). What was a mess when you looked at
> > > it? (Just in case I'm missing something.)
> >
> > it seems to be a fairly random mix of monospaced and proportional text
> > for me. for example, the first line "Index:" is proportional, but then
> > the --- and +++ lines are monospaced, and it goes back and forth a
> > lot. (in both Chrome and Firefox.)
> >
> That's because the HTML code produced by Lurker is wrong.
> E.g. a text around CHECK_INCLUDE_FILE line reads:
>
> <br> INCLUDE(CheckTypeSize)
> <br> <br><pre class="art"> CHECK_INCLUDE_FILE(dirent.h     HAVE_DIRENT_H)

>
> It seems that Lurker presents the plain text messages with a proportional text,
> but when it thinks a piece of text looks like a code, it switches to an
> unproportional block. And it cannot recognize the end of the patch properly.
>
> This is either a bug in ther Lurker, or the e-mail was sent erroneously with
> a format=flowed MIME attribute.


if gmail's "Show original" is to be believed, the patch wasn't
format=flowed, so it's presumably just lurker's problem (which would
explain why the downloaded patch is okay).

and to think that i thought that just sending a patch to the mailing
list would be less error-prone than trying to use the bug database...
:-)

> -- Petr
>
> --
> ## List details at https://lists.exim.org/mailman/listinfo/pcre-dev