From: Johannes Sixt <j6t@kdbg.org>
Date: Tue, 20 Dec 2016 20:12:27 GMT
Subject: Re: [PATCHv4 3/5] run-command: add {run,start,finish}_command_or_die
Message-ID: <f14ee492-8297-c8ec-f80f-f8f24caf91e1@kdbg.org>
In-Reply-To: <CAGZ79kYNKWfnEXWJfyRUutFyaQiRD9qW--LkK4Nbwdf7FtdPQA@mail.gmail.com>

Am 20.12.2016 um 20:23 schrieb Stefan Beller:
> In a reroll I'll drop this patch and instead introduce
> a function `char *get_child_command_line(struct child_process*);`, which
> a caller can call before calling finish_command and then use the
> resulting string
> to assemble an error message without lego.

That sounds a lot better, thank you. Note though, that the function 
would have to be called before start_command(), because when it fails, 
it would be too late.

I have to ask, though: Is it so much necessary to report the exact 
command line in an error message? If someone is interested in which 
command failed, it would be "sufficient" to run the command under 
GIT_TRACE=2.

-- Hannes