Re: New built-in fsmonitor: messages accidentally on stdout?

From: Tao Klerks <tao@klerks.biz>
Date: Sep 27, 2021, 09:32 UTC
Message-ID: <CAPMMpoic0Dt=5Z7GB7BG7CqwxAs49SPPDkdasUPvasvsPhr-tg@mail.gmail.com>
In-Reply-To: <CAPMMpoj-oiD1yDHx6WS_1uiPOXMLj7FcfBPZm-q4vO30kpE_ow@mail.gmail.com>

Hi folks,

I did track down where that message was being issued onto stdout in git-for-windows (https://github.com/git-for-windows/git/blob/main/builtin/fsmonitor--daemon.c), and I see that in the the latest version of the patch series it was already switched to stderr (https://github.com/jeffhostetler/git/blob/f08a63e197a0efe4d8f0452893238b0d3156e669/builtin/fsmonitor--daemon.c).

The only remaining thing I can ask for, I guess, is that the next Git for Windows release ideally include this change/fix, even though the built-in fsmonitor is not in "regular git" at all yet?

Thanks, Tao

On Fri, Sep 24, 2021 at 8:02 PM Tao Klerks <tao@klerks.biz> wrote:

Show 22 quoted lines

>
> Hi folks,
>
> I've been testing the new "core.usebuiltinfsmonitor=true" capability
> in Git for Windows 2.33.0.windows.2, and came across something today
> that threw off my scripting:
>
> In general, git "informational" messages are always output on stderr;
> this caught me out initially, but is a godsend for scripting.
>
> However, with this new functionality, the "starting fsmonitor-daemon
> in 'C:/whatever" message is being output on stdout!
>
> This interferes with scripting around things like "git status --porcelain".
>
> I understand this fsmonitor stuff is still somewhat in-flux (patched
> into the windows port but not yet "regular" git releases?), and this
> issue may already have been reported and addressed in the latest set
> of patches - my apologies if that is the case!
>
> Best regards,
> Tao