git/list[1] front-page[2] threads[3] people[4] search[5] about
 

RE: preparing for 2.34.1

From
rsbecker@nexbridge.com <rsbecker@nexbridge.com>
Date
Nov 22, 2021, 23:51 UTC
Message-ID
<039401d7dffb$e5198800$af4c9800$@nexbridge.com>
In-Reply-To
<nycvar.QRO.7.76.6.2111230025440.63@tvgsbejvaqbjf.bet>
On November 22, 2021 6:30 PM, Johannes Schindelin wrote:
Show 44 quoted lines
> On Mon, 22 Nov 2021, Junio C Hamano wrote:
> 
> > Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> >
> > > The quickest workaround for this is probably to special-case the
> > > editor
> > > `echo`:
> >
> > "GIT_EDITOR=: git cmd" would also be a common trick people would use
> > to bypass editor and take whatever is given as an initial template.
> 
> GIT_EDITOR=: is not a problem because of
> https://github.com/git/git/blob/v2.34.0/editor.c#L59:
> 
> 	if (strcmp(editor, ":")) {
> 		[...]
> 		term_fail = save_term(1);
> 		if (start_command(&p) < 0) {
> 			if (!term_fail)
> 				restore_term();
> 			[...]
> 		}
> 
> 		[...]
> 		if (!term_fail)
> 			restore_term();
> 		[...]
> 	}
> 
> > > However, I could imagine that other scenarios call for an editor
> > > that _also_ does not run in the terminal, and where also no real
> > > terminal is available for saving and restoring.
> > >
> > > I was tempted to suggest an `isatty(2)`, but that probably comes
> > > with its own problems, too.
> >
> > I think isatty(2) is pretty much our synonym to "are we talking to an
> > end-user sitting in front of the terminal".  Mostly we use it as a way
> > to control the progress bars, and use of editor on terminal would be
> > in line with these existing uses.
> 
> Indeed, I think that isatty(2) is a better indicator than isatty(1). We
> sometimes _do_ redirect the output of, say, `git commit`, to capture the
> commit hash that was generated. We typically do not redirect stderr,
though,
> unless calling from an application and capturing everything via pipes. So
> isatty(2) strikes me as the best balance we can strike here.

Please be careful of this one. isatty(2) may not be the issue, but the filedes provided by Jenkins and other CI/CD systems over SSH often mischaracterises the results from this call. Hacking the file descriptor (2>&1 etc) prior to going to git is a common thing in scripts. Stdin is frequently wrong when Jenkins sets up the environment to prompt for an SSH passphrase - happens in non-Docker nohup situations - where git will end up thinking it is interactive when it is not - not on NonStop fortunately, but this happens with a Gentoo Hypervisor and Ubuntu VM. Also, stderr gets redirected in scripts frequently in my experience - particularly on exotic operating systems, crossing over from say, legacy NonStop or IBM TSO to each's POSIX environment. I would expect breakages from this assumption, so please be cautious.

-Randall
Previous: Johannes SchindelinNext: Carlo Arenas
Message 6 of 10 in “preparing for 2.34.1”
  1. Junio C HamanoNov 22, 2021
  2. Carlo ArenasNov 22, 2021
  3. Johannes SchindelinNov 22, 2021
  4. Junio C HamanoNov 22, 2021
  5. Johannes SchindelinNov 22, 2021
  6. rsbecker@nexbridge.comNov 22, 2021
  7. Carlo ArenasNov 22, 2021
  8. Derrick StoleeNov 22, 2021
  9. Junio C HamanoNov 22, 2021
  10. preparing for 2.34.1 (2nd round)Junio C Hamano, Nov 23, 2021

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.