Hi Michael!

I have cloned the git repo and tested your commit.  It works well in my 
environment!

> ...but you shall be aware that the history file contains also the
> commands Tramp has sent in the background.

I see.  It is reasonable and acceptable.

> You shall also set tramp-histfile-override to nil in order to get the
> expected history file.

OK, thanks for reminding~

Fan


> 在 2019年7月22日，19:11，Michael Albinus <michael.albinus@gmx.de> 写道：
> 
> Fan Yang <fan_yang@sjtu.edu.cn <mailto:fan_yang@sjtu.edu.cn>> writes:
> 
> Hi Fan,
> 
>> * Bug description:
>> 
>> When using "M-x shell" on a remote file (say "/scp:yf@val17:"), a bash
>> is started at the server "val17" as the user "yf".  But the "*shell*"
>> buffer is started with an empty input ring.  It does not read bash
>> history file ("/scp:yf@val17:~/.bash_history") into comint input ring.
>> 
>> * What is expected:
>> 
>> When using "M-x shell" on a remote file (say "/scp:yf@val17:"), a bash
>> is started at the server "val17" as the user "yf".  The "*shell*"
>> 1. has the content of "/scp:yf@val17:~/.bash_history" as comint input
>>   ring.  And I can use "M-p" to insert previous command, use "M-r" to
>>   isearch previous comamnd, etc.
>> 2. the commands I have input in this "*shell*" will be appended to
>>   "/scp:yf@val17:~/.bash_history" after shell exit.
>> Just like what "shell-mode" does for local shell.
> 
> I've fixed this in the Emacs repository, commit 7f95d2d407. It works as
> expected, but you shall be aware that the history file contains also the
> commands Tramp has sent in the background.
> 
> You shall also set tramp-histfile-override to nil in order to get the
> expected history file.
> 
> Do you have a chance to test with a recent Emacs 27.0.50?
> 
>> Best Regards,
>> Fan
> 
> Best regards, Michael.