git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Shawn O. Pearce" <spearce@spearce.org>
To: "Roger C. Soares" <rogersoares@intelinet.com.br>
Cc: git@vger.kernel.org, robin.rosenberg@dewire.com
Subject: Re: [EGIT PATCH 1/4] Change history page table to SWT.VIRTUAL.
Date: Mon, 31 Mar 2008 23:36:14 -0400	[thread overview]
Message-ID: <20080401033614.GP10274@spearce.org> (raw)
In-Reply-To: <47F1AB1B.90309@intelinet.com.br>

"Roger C. Soares" <rogersoares@intelinet.com.br> wrote:
> Shawn O. Pearce escreveu:
> >Yea, I originally wrote my series around the VIRTUAL flag but on
> >Win32 it caused ArrayIndexOutOfBoundsExceptions to be thrown from
> >deep down within the Win32 implementation of the SWT Table widget.
> >
> >Appears to be something of a known bug, based on the Eclipse issue
> >tracker, but not much work happening to fix it.
>
> Hum, a VIRTUAL table sounds like a very usefull feature to be badly 
> broken on windows, at least there should be some workaround... is it 
> easily reproducible?

I tested your patch this evening on Win32.  Its nearly instant.
Seriously.  gitk can't touch it on the same system.  The bug I
was seeing didn't happen.

Originally I wrote the graph table using the ILazyContentProvider
*and* the SWT.VIRTUAL flag.  I didn't realize you had the option to
set only the SWT.VIRTUAL flag.  I think the bug on Win32 is related
to how the ILazyContentProvider gets used by the JFace TableViewer
and less about the SWT.VIRTUAL flag itself.

So anyway, I see no breakage on Mac OS X or Win32 with your patch,
and you said you tested it on Linux, so I'm going to include it.
Thanks for figuring that one out, its a nice performance boost.

> >I'll retest this tomorrow on Win32, but I'm pretty certain its
> >a bad idea on that platform.  What are you running on, Linux?
> >Maybe we can set this flag everywhere except on Win32
>
> Yep, linux.
> 
> Maybe another option to try before leaving windows out is the 
> ILazyContentProvider. Have you noticed that while GenerateHistoryJob is 
> updating the table you can't use it? Because the input is regenerated 
> every time, the table keeps going back to the first row.

Hmm.  I didn't notice that, but at this point its so damn fast for
me that I don't have the reflexes to really try and use the table
before GenerateHistoryJob is complete.  I'll have to add some sleeps
in there to make it slow down its work and see if I can reproduce
what you are describing.

-- 
Shawn.

  reply	other threads:[~2008-04-01  3:37 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-03-30 15:18 [EGIT PATCH 1/4] Change history page table to SWT.VIRTUAL Roger C. Soares
2008-03-31  5:34 ` Shawn O. Pearce
2008-04-01  3:25   ` Roger C. Soares
2008-04-01  3:36     ` Shawn O. Pearce [this message]
2008-04-01  3:54       ` Roger C. Soares
2008-04-01  4:12         ` Shawn O. Pearce

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080401033614.GP10274@spearce.org \
    --to=spearce@spearce.org \
    --cc=git@vger.kernel.org \
    --cc=robin.rosenberg@dewire.com \
    --cc=rogersoares@intelinet.com.br \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).