Re: [RFC PATCH 3/3] FAQ: sync newuser-install

[Eric Cook <llua@gmx.com>]

On 8/24/22 15:11, Felipe Contreras wrote:
> On Wed, Aug 24, 2022 at 1:33 PM Eric Cook <llua@gmx.com> wrote:
>>
>> On 8/24/22 13:32, Felipe Contreras wrote:
>>> On Wed, Aug 24, 2022 at 4:12 AM Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
>>>>
>>>> Felipe Contreras wrote on Wed, 24 Aug 2022 04:31 +00:00:
>>>>> Better to use the same HISTFILE in both.
>>>>
>>>> Agreed, but...
>>>>
>>>>> +++ b/Etc/FAQ.yo
>>>>> @@ -1629,7 +1629,7 @@ label(321)
>>>>>      In zsh, you need to set the location where you want the history to be written
>>>>>      with tt($HISTFILE).  For example,
>>>>>      verb(
>>>>> -    HISTFILE=~/.zsh_history
>>>>> +    HISTFILE=~/.histfile
>>>>
>>>> The default name should include "zsh" somewhere for coexistence with
>>>> other programs.  So, I'd rather keep the FAQ as-is.
>>>
>>> I agree having a '.zsh_' prefix is more consistent with other
>>> programs, but then that's what should be in the newuser-install
>>> script, no?
>>>
>>> I actually think this should be the default, no need for *everyone* to
>>> specify this.
>>
>> Yes *everyone* should specify it.
>
> Why?
>
> All programs dump their history by default. Node, python, ruby, less,
> mariadb, and of course bash and ksh. Those are the ones that I can
> easily find.
>To avoid the problem i mentioned, also there are existing users that like
the current behavior since they do not use shell history. at work i don't use it.

>> Aside from the arbitrary bump in HISTSIZE, if HISTFILE is set by default and
>> zsh is invoked with -f which is common when debugging or testing you get the
>> expected defaults. But if the user typically increases SAVEHIST from said
>> default, because regardless of what you set it to, people will change it to
>> their preference, you create the same problem that other shells like bash have,
>> the truncation of HISTFILE to the default SAVEHIST upon the exit of that zsh -f session.
>> usually without the user noticing for a period of time.
>
> So because of this corner case that 99.99% of people are never going
> to hit we should bother 100% of users. It doesn't seem like an issue
> worthy of so much consideration. Moreover, it could be solved in other
> ways, like not setting HISTFILE when NO_RCS is on.

Starting the shell without your dotfiles is commonly given advice when troubleshooting
weird behavior, found across this mailing list, places like stackexchange and irc,
It's not a corner case.