Re: [PATCH 0/2] fix http deadlock on giant ref negotiations
From: Dennis Kaarsemaker <hidden>
Date: 2016-06-15 23:04:47
On vr, 2015-05-15 at 04:38 -0400, Jeff King wrote:
On Fri, May 15, 2015 at 09:41:20AM +0200, Dennis Kaarsemaker wrote:quoted
quoted
I have no clue how to write a test that would trigger this reliably without requiring a gigantic test fixture. However, I did confirm that it fixes the problem on the chromium case you provided (which otherwise deadlocks reliably for me).This looks similar to the failure I posted about alst year in http://thread.gmane.org/gmane.comp.version-control.git/258514 Though the issue is different, it has the same 'hanging git fetch' symptom due to the deadlock between upload-pack and http-backend.Thanks, I think it is the same issue (in the end I was replicating not with `--reference`, but just by doing a fetch from the other repository). And our solutions are essentially the same. I do prefer mine because: 1. It keeps the buffering logic in http-backend; the half-duplex nature is an http detail. 2. I think it's better to buffer the request rather than the response, for the reasons I stated in the commit message.quoted
The patch I sent back then is suboptimal, as it can cause larger packs than necessary (we still use it though, as the alternative is a non-working git), but it does include a test you may be able to use to verify your fix, if this is indeed the same issue.I applied the test from your patch, but couldn't get it to fail even with stock git. The test above it shrunk a bit, but I was able to tweak yours to generate tags from 2001..100000, which I thought would have worked. I suspect it's something silly like the size not being quite big enough for the pipe buffer on my system, or something like that. Though I couldn't get it to fail even with 200,000 tags, so perhaps it's something else.
The shrinkage in the test above it will actually work around the issue, as there are now fewer already-fetched tags to negotiate. Either reverting that shrinkage or executing the new test twice should do the trick. -- Dennis Kaarsemaker http://www.kaarsemaker.net