Leo Liu <sdl.web@gmail.com> writes:

> On 2018-06-12 08:21 +0300, Eli Zaretskii wrote:
>> Ugh, macOS...  This means no xbacktrace and no other GDB
>> wizardry...
>
> Indeed.
>
>> Well, I think the most important thing is to produce a Lisp
>> backtrace from the error.
>
> Turns out 26.1 cannot load a version of python.el (by dave love) that I
> have in my site-lisp.
>
> 1. emacs -Q -l python.el
> 2. M-x python-mode
>
> ,----[ error ]
> | Variable binding depth exceeds max-specpdl-size
> | Error in post-command-hook (global-eldoc-mode-check-buffers): (error "Variable binding depth exceeds max-specpdl-size")
> `----
>
> Gnus was just loading python-mode to fontify some code in an article.

double posting since I sent the email only to Leo.

VVVV original email VVVV

That seems to be a result of converting `global-eldoc-mode' to use
`define-globalized-minor-mode' in 2349f. The recursive call seems to be
a result of the following sequence of calls:

1. python-mode is enabled which adds an eldoc-mode-hook
2. eldoc-mode is turned on in the buffer triggering python's
eldoc-mode-hook
3. python-mode hook will start an inferior process which in turn trigger
,MODE-enable-in-buffers
4. ,MODE-enable-in-buffers will go over the list again trying to enable
eldoc-mode

The problem is in step 4. The eldoc-global-mode buffer list isn't reset
in step 2. Step 4 will try to enabe the mode for the same buffer and
start a sequence of calls at 2. I was able to fix this problem by
setting the buffer-list to nil inside ,MODE-enable-in-buffers. I
attached a patch below.