Re: Recommend these .gitconfig settings for git integrity.

[Paul Eggert]

On 02/01/2016 12:28 PM, Eli Zaretskii wrote:
> It's a general security-related issue

As I understand things it's more general than that. Setting transfer.fsckObjects 
also catches things like transmission and storage errors.

Should we pay the checking overhead to catch these reliability and security 
problems? That's the question. Emacs developers who are using this setting 
report that the overhead is small enough as to not be a significant issue, which 
suggests that the extra checking is a good idea.

> that Emacs shouldn't force on its users.

The script is intended to be run only by Emacs developers, and is optional even 
for developers. It's certainly not forcing anything on Emacs users.


On 02/01/2016 01:40 PM, Stefan Monnier wrote:
> At least not before we lobby the Git crowd to turn it on by default.

Any such lobbying effort could well fail, as the checking overhead can be 
substantial in some projects, which means that changing the default for all 
Git-using projects could well be a bad idea.

Besides, if the idea is a good one for Emacs development, then even if we 
successfully lobby for it to be the Git default it will take quite some time for 
the change to propagate, and in the meantime we win by setting the Git 
configuration appropriately for Emacs. Conversely, if our lobbying effort fails, 
we still win by setting the Git configuration appropriately for Emacs now.


On 02/01/2016 12:34 PM, Óscar Fuentes wrote:
> The script I use for building Emacs contains
>
> git clean -f -d -x
> ./autogen.sh

Surely this script is unaffected by the benign git config settings that 
autogen.sh installs. However, as you're worried about the possibility, I added 
support for a new option to autogen.sh, so if you change that script's autogen 
line to:

   ./autogen.sh --git-config=false

then autogen.sh won't use "git config" and will not install the setting that 
started this thread.

> How about mentioning (not applying) the recommended settings on autogen.sh

That would make autogen.sh less convenient to use. I largely ignore autogen.sh's 
output, and I expect it to work without my having to fiddle with Git 
configuration or hooks by hand afterwards. Besides, autogen.sh has been 
configuring Git for quite some time with no practical problems reported; why 
change something that isn't broken?

> How those settings benefit Emacs development?

Assuming you're asking about the recent change, there is some justification for 
it in the first part of this email.