<div> </div>
<div>Hi, Martin.</div>
<div> </div>
<div>Your observation about phones is certainly correct. I would ask how this scales on the bzr smartserver end of the wire. If, say, four developers are merging to/from four different branches [1] simultaneously, it seems like the box running the bzr smartserver might thrash badly.</div>
<div> </div>
<div>What do you expect?</div>
<div> </div>
<div>[1] For simplicity, assume plain branches. For complexity, assume four branches in a shared repo.</div>
<div> </div>
<div>Thanks </div>
<div>~M<br><br></div>
<div class="gmail_quote">On Fri, Jan 21, 2011 at 12:58 PM, Martin Pool <span dir="ltr"><<a href="mailto:mbp@canonical.com">mbp@canonical.com</a>></span> wrote:<br>
<blockquote style="BORDER-LEFT: #ccc 1px solid; MARGIN: 0px 0px 0px 0.8ex; PADDING-LEFT: 1ex" class="gmail_quote">
<div class="im">On 14 January 2011 22:51, Doug Lee <<a href="mailto:dgl@dlee.org">dgl@dlee.org</a>> wrote:<br>> I believe this means I have a repository that can't work in less than<br>> 512 meg of physical memory, which is quite a requirement.<br>
<br></div>Since they now sell phones with more than that 512MB memory I don't<br>know if it's really a lot anymore. But it certainly is an annoying<br>situation if that's what the machine you have has.<br>
<div class="im"><br>> Further advice welcome and sorry for the length of this thread.<br><br></div>We still have some things where the memory usage grows with the size<br>of the last file: I think that we no longer have anything that scales<br>
with the pack size or the total tree size. (Or maybe some network<br>stream operations are still in that class.)<br><br>The main thing I would suggest is getting to the tip of bzr and bzr-svn.<br><br>You could try adding more swap space; it will be slow but it may complete.<br>
<br>--<br><font color="#888888">Martin<br><br></font></blockquote></div><br>