git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Hector Santos <hsantos@isdg.net>
To: git@vger.kernel.org
Subject: Git Daemon on Windows fatal error.
Date: Wed, 31 May 2017 15:10:50 -0400	[thread overview]
Message-ID: <592F153A.7020403@isdg.net> (raw)

Hi, I am relatively new to GIT (coming from CVS and SVN) and I am 
trying to setup "Git Daemon" on windows.

I got it working for Local network communications:

d:\local\wc5\testgit>git clone git://localhost/http clone10
Cloning into 'clone10'...
remote: Counting objects: 526, done.
remote: Compressing objects: 100% (520/520), done.
Receiving objects: 100% (526/526), 1.38 MiB | 0 bytes/s, done.
remote: Total 526 (delta 81), reused 0 (delta 0)
Resolving deltas: 100% (81/81), done.

but it fails over the wire when using the public host domain:

d:\local\wc5\testgit>git clone git://public.example.dom/http clone11
Cloning into 'clone11'...
remote: Counting objects: 526, done.
remote: Compressing objects: 100% (520/520), done.
remote: Total 526 (delta 81), reused 0 (delta 0)
fatal: read error: Invalid argument
fatal: early EOF
fatal: index-pack failed

Sometimes its a different initial fatal error but generally the same.  
Once or twice, a repeat MAY work, but often not.

Short of digging into the git source code, I did as much research 
online and tried the various config options suggestions, changing the 
packet size, etc, to no avail.

To me, this seems like a "Socket Half Close" problem.   If anyone is 
aware of what appears to be a long time "known" problem, and have a 
real solution, it would be greatly appreciated.   Otherwise, I am very 
interesting in exploring the Half Close solution as I've seen similar 
behavior in other internet hosting servers in the past.   A simple 
closesocket() wrapper funciton did the trick:


// HalfCloseSocket() performs a TCP Half Close by calling shutdown()
// which signals the remote that no more data is going to be
// sent (FIN signal). HalfCloseSocket() then goes into a
// recv() loop to wait for the remote to acknowledge the close.
// This acknowledgment comes as a recv() return value
// of zero (less).

BOOL HalfCloseSocket(SOCKET socket)
{
     if (shutdown(socket,SD_SENT) != 0) {
         return FALSE;
     }
     int ret = 0;
     int msecs = 10; // poor man sanity check
     char buf[8*1024];
     while ((ret = recv(socket, buf,sizeof(buf),0)) > 0) {
         buf[0] = 0;
         buf[1] = 0;
         msecs--;
         if (msecs == 0) break;
     }
     return closesocket(socket);
}

While I rather not get into the source, I am willing to explore the 
effort if there is no other option.

Thanks for any input you can provide

Hector Santos
Santronics Software, Inc.

-- 
HLS




             reply	other threads:[~2017-05-31 19:16 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-05-31 19:10 Hector Santos [this message]
2017-06-01  6:53 ` Git Daemon on Windows fatal error Torsten Bögershausen

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=592F153A.7020403@isdg.net \
    --to=hsantos@isdg.net \
    --cc=git@vger.kernel.org \
    /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).