Re: Git clone failure

From: Konstantin Khomoutov <kostix@bswap.ru>
Date: Sep 24, 2024, 14:20 UTC
Message-ID: <20240924142046.6wkhwnfd5vp5wh3y@carbon>
In-Reply-To: <SI2P153MB071833367C62DD94C96CF1F7D46F2@SI2P153MB0718.APCP153.PROD.OUTLOOK.COM>

On Mon, Sep 23, 2024 at 10:47:08AM +0000, Krishna Vivek Vitta wrote:

Show 7 quoted lines

> We've seen multiple reports of git repositories failing to clone / getting
> corrupted in p9 file system. The mount points under this file system are
> marked for FANOTIFY to intercept file system events
> 
> When we remove the marking on these mount points, git clone succeeds.
> 
> Following is the error message:

[...]

> fatal: unknown error occurred while reading the configuration files
> 
> Any reason why it is failing?

[...]

Are you able to build Git from the source?

As you can see in config.c, the error is reported in repo_read_config in the case the function config_with_options, it calls, fails. Since it's not expected to fail - except for some weird reasons, - you're probably dealing with such a reason, and so you'd probably need to debug it yourself because I hardly beleive it's easily possible for someone else to recreate your specific setup. The debugging might be a simple "printf-style" one - that is, navigate the call chain of these configuration-reading functions, stick calls to printf() in places where some nested function returns a value indicating an error exit, build, run, and see which of these printf() calls pinpoints the failure. You will probably trace this to call to some C-library I/O function.