<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
<title></title>
</head>
<body text="#000000" bgcolor="#ffffff">
On 11/24/2010 12:34 PM, Thierry Crozat wrote:
<blockquote
cite="mid:AANLkTikOR90Hi_QTS_BuL5SfFKZPfjvYewadhR-OY0+6@mail.gmail.com"
type="cite">Hi,<br>
<br>
<div class="gmail_quote">On 24 November 2010 10:10, Max Horn <span
dir="ltr"><<a moz-do-not-send="true"
href="mailto:max@quendi.de">max@quendi.de</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid
rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left:
1ex;">
Am 24.11.2010 um 01:41 schrieb Johannes Schickel:<br>
<div class="im"><br>
> 1) Log strategy<br>
><br>
> We can basically have two different logging strategies:<br>
><br>
> a) only have one log file, which is overwritten on
every run.<br>
> b) have a fixed maximum of log files and always reuse
the oldest one on startup (i.e. log rotation).<br>
> c) have an unlimited number of log files.<br>
<br>
</div>
I am all for (a) at this point, and will quote myself from a
tracker comment:<br>
<br>
Right now I would prefer to go with the "single log file"
approach. A single log file is very easy to understand for the
user, and there is little risk of the user picking "the wrong
one". It is also easiest to implement for us. E.g. no need to
rotate log files to prevent the disk from filling up with lots
of "-d9" log files.<br>
<br>
Potential drawback is that the user must remember to copy it
right away, before re-starting ScummVM, else it is lost. This
is mostly a problem for "rare" hard-to-reproduce issues. If
the user can easily reproduce it, then they are just fine
either way. Moreover, if we crash into the debugger console,
we could display a text there like "If you want to file a bug
report on this, you can find a log file with relevant info in
location XYZ. Please copy it NOW, before restarting ScummVM,
otherwise it will get overwritten." Well, something like that,
at least.<br>
<br>
</blockquote>
</div>
It seems given here that one log file corresponds to one run of
ScummVM, yet I have seen no discussion of this. Many applications
(on Linux at least, I am not very familiar with other systems) use
a single log file and append logs to them with a timestamp when
restarting the application. This files can get very big but we
could improve a bit the system by overwritting the file if for
example the creation date is not today (i.e. all the runs for a
day are stored in the log file and the files get overwritten the
next day). For the user this would still be simple (only one log
file) and remove some of the problem of overwritting the log file
when restarting ScummVM. One drawback is that this file may still
get big (especially when making several run with -d9, but maybe if
the user knows how to start an application with -d9 he would not
mind to much having to manually delete the old log file if he
wants to have a clean log for a new run).<br>
</blockquote>
<br>
<br>
Another drawback is that if the user submits a log it might not be
clear which of the embedded "sub" logs is actually the one from the
bug report, especially if he does another run after he encountered
the bug first.<br>
<br>
<br>
<blockquote
cite="mid:AANLkTikOR90Hi_QTS_BuL5SfFKZPfjvYewadhR-OY0+6@mail.gmail.com"
type="cite">
<br>
I don't know if this is a good idea, but just wanted to say that
there may be more options than just the three proposed by
Johannes.<br>
</blockquote>
<br>
As you might have noticed that first sentence shouldn't be there
either, since it does only talk about "two" different strategies
aye. But yeah there are other alternatives there and thanks for
bringing this one up.<br>
<br>
// Johannes<br>
</body>
</html>