Thread: [Madwifi-devel] x-large svn-changesets
Status: Beta
Brought to you by:
otaku
From: Paolo <oo...@us...> - 2006-06-15 15:39:08
|
hi, I was upgrading my local madwifi-ng codebase via changeset-*.diff through a one-liner script, then got surprise: 7203022 Jun 14 21:05 madwifi_changeset-1532.diff 8924923 Jun 8 09:27 madwifi_changeset-1631.diff 28782174 Jun 8 09:26 madwifi_changeset-1606.diff and .zip are alsmost same size, except 1532 : 1649467 Jun 15 17:31 madwifi_changeset-1532.zip 8924916 Jun 8 09:27 madwifi_changeset-1631.zip 28782167 Jun 8 09:26 madwifi_changeset-1606.zip Is there anything wrong here? I mean, 28MB of diff/zip whereas full snapshot is just ~1MB ?!? thx -- paolo |
From: Paolo <oo...@us...> - 2006-06-15 15:45:14
|
On Thu, Jun 15, 2006 at 05:38:54PM +0200, Paolo wrote: > 28782167 Jun 8 09:26 madwifi_changeset-1606.zip > > Is there anything wrong here? I mean, 28MB of diff/zip whereas full snapshot > is just ~1MB ?!? 3.5 MB actually - still weird though. -- paolo |
From: Yasen B. <fra...@gm...> - 2006-06-15 16:13:12
|
That's normal. If you have 1MB source v1 and completely different 1MB source v2, diff will be ~2MB diff is less than the whole source only when you have a near version to start with. 2006/6/15, Paolo <oo...@us...>: > On Thu, Jun 15, 2006 at 05:38:54PM +0200, Paolo wrote: > > 28782167 Jun 8 09:26 madwifi_changeset-1606.zip > > > > Is there anything wrong here? I mean, 28MB of diff/zip whereas full snapshot > > is just ~1MB ?!? > > 3.5 MB actually - still weird though. > > -- paolo > > > > _______________________________________________ > Madwifi-devel mailing list > Mad...@li... > https://lists.sourceforge.net/lists/listinfo/madwifi-devel > |
From: Paolo <oo...@us...> - 2006-06-15 20:03:15
|
On Thu, Jun 15, 2006 at 07:13:05PM +0300, Yasen Balev wrote: > That's normal. > > If you have 1MB source v1 and completely different 1MB source v2, diff > will be ~2MB hm, right, at most it'd ~double, assuming not too many hunks. Though I wouldn't call 'normal' a ~3 x codebase size diff (codebase is ~9.5MB), uncompressible, for a changeset 1605->1606. Indeed, now that got a chance to inspect those diffs, it looks like Trac didn't honour format=diff request, nor the format=zip ones, as it seems what I got is the same fully indexed [XH]TML changeset page in both cases - which crashed FF and torn down wmaker, wdm and X along w/ a dozen other xterm :P : % less madwifi_changeset-1606.zip <!DOCTYPE html ... </tr></tbody><tbody><tr><th>106</th><th>106</th><td class="l"><span>M"CP0! .Y'$03L1PX`7R(3!.M'`0"?(AO^7],.`"HL+@`I(4$"*4A@!PE(``0!</span> </td></tr><t r><th>107</th><th>107</th><td class="l"><span>M1`X`"CP-,*!!```LH*$-+4#(_S_D``!&g t;IP@`/J40`%ZE&`!^I2``GJ4H`+ZE</span> </td></tr></tbody><tbody class="m od"><tr class="first"><th>108</th><th> </th><td class="l">M,`#>I4``WB,!@ /I ah, that was 'Update HAL to version 0.9.17.0' - perhaps that's a case where (indexed) diff hurts more than it serves. -- paolo |
From: Michael R. <ma...@no...> - 2006-06-16 03:52:04
|
Hi. Paolo wrote: > 7203022 Jun 14 21:05 madwifi_changeset-1532.diff "Decode HAL files from uuencoded text to binaries ..." > 8924923 Jun 8 09:27 madwifi_changeset-1631.diff "Update HAL to version 0.9.17.2" > 28782174 Jun 8 09:26 madwifi_changeset-1606.diff "Update HAL to version 0.9.17.0 ..." As you can see, the size of these diffs is caused by changes regarding the HAL. Since the HAL object files are uuencoded (making them appear as non-binary files), Subversion doesn't mask changes to these files when diffing. And even simple changes to the HAL objects produce a bunch of lines in the diffs. Bye, Mike |
From: Paolo <oo...@us...> - 2006-06-16 15:02:18
|
On Fri, Jun 16, 2006 at 05:52:02AM +0200, Michael Renzmann wrote: > > 28782174 Jun 8 09:26 madwifi_changeset-1606.diff > > "Update HAL to version 0.9.17.0 ..." > > As you can see, the size of these diffs is caused by changes regarding > the HAL. Since the HAL object files are uuencoded (making them appear as yes, but as mentioned the problem turned out to be that Trac didn't serve format=diff as .diff nor format=zip as .zip - always the annotated webpage, which surely is much bigger than both plain .diff and .zip. For some reason - perhaps just the size - it did not honour the format request. I guess this is not the intended behaviour. -- paolo |
From: Michael R. <ma...@no...> - 2006-06-20 03:30:09
|
Hi. Paolo wrote: >> As you can see, the size of these diffs is caused by changes regarding >> the HAL. Since the HAL object files are uuencoded (making them appear as > yes, but as mentioned the problem turned out to be that Trac didn't serve > format=diff as .diff nor format=zip as .zip - always the annotated webpage, > which surely is much bigger than both plain .diff and .zip. > For some reason - perhaps just the size - it did not honour the format > request. I guess this is not the intended behaviour. I have an explanation for this behaviour: Trac's changeset viewer causes quite some load on the server for large changesets, most notably those which import new versions of the HAL. Some time back I disabled viewing these changesets with some rewrite rules in Apache's configuration, but meanwhile I redirect these requests to statical "copys" of the resulting pages. This way users still see the desired result, but the server is much more relaxed. The only downside of this is that the format=x "parameters" are not honoured so far, and this is the reason you always get the HTML file instead of the desired format. As a "workaround" you should use Subversion to get the diffs for now: svn diff -r 1605:1606 http://svn.madwifi.org > changeset1606.diff This works at least with svn v1.1.4 (just tested). Bye, Mike |
From: Paolo <oo...@us...> - 2006-06-20 21:39:08
|
On Tue, Jun 20, 2006 at 05:29:59AM +0200, Michael Renzmann wrote: > > For some reason - perhaps just the size - it did not honour the format > > request. I guess this is not the intended behaviour. > > I have an explanation for this behaviour: Trac's changeset viewer causes ... > much more relaxed. The only downside of this is that the format=x > "parameters" are not honoured so far, and this is the reason you always > get the HTML file instead of the desired format. thanks for the ex. Wouldn't be better to only allow for .diff and/or .zip? just wondering who's gonna read through some klines of annotated binary diffs. -- paolo |
From: Michael R. <ma...@no...> - 2006-06-21 03:59:13
|
Hi. Paolo wrote: >> much more relaxed. The only downside of this is that the format=x >> "parameters" are not honoured so far, and this is the reason you always >> get the HTML file instead of the desired format. > thanks for the ex. > Wouldn't be better to only allow for .diff and/or .zip? just wondering > who's gonna read through some klines of annotated binary diffs. The changeset browser as such is quite helpful in daily work. In addition, when a new HAL version is imported, chances are that also files are modified which are not uuencoded. They are huge changesets, but allowing them to be browsed still makes sense, IMO. There is no way to differentiate between requests with and without format-parameter. The "redirection" technically is solved like this: === cut === <Location /changeset/1631> SetHandler None </Location> Redirect permanent /changeset/1631 http://madwifi.org/changeset_static/1631.html === cut === I have to disable Trac handling for ALL requests (including those for diffs or zips) for a given changeset - no exceptions can be made. For now I don't have the time to work around this, sorry. Bye, Mike |