Re: guc: make dereference style consistent in check_backtrace_functions - Mailing list pgsql-hackers

From Chao Li
Subject Re: guc: make dereference style consistent in check_backtrace_functions
Date
Msg-id 08805314-299A-4941-B63E-4CC3E144E0E7@gmail.com
Whole thread Raw
In response to guc: make dereference style consistent in check_backtrace_functions  (zhanghu <kongbaik228@gmail.com>)
Responses Re: guc: make dereference style consistent in check_backtrace_functions
List pgsql-hackers

> On Feb 26, 2026, at 15:03, zhanghu <kongbaik228@gmail.com> wrote:
>
> Hi,
>
> In check_backtrace_functions(), most accesses to the input string follow the pattern (*newval)[i]. However, the
empty-stringcheck is currently written as: 
>
> if (*newval[0] == '\0')
>
> While functionally correct due to how the compiler handles the address-of-address context here, this form is
semanticallymisleading. It relies on implicit operator precedence rather than explicit intent. 
>
> The attached patch rewrites it as:
>
> if ((*newval)[0] == '\0')
>
> This change ensures semantic clarity and maintains a consistent dereferencing style throughout the function. No
functionalchanges are introduced. 
>
> Regards,
> Zhang Hu
> <v1-0001-guc-make-dereference-style-consistent-in-check_ba.patch>

This is an interesting find.

[] has higher precedence than *, so:

 - (*newval)[i] means to get the first string, then get the char at position i
 - *newval[i] means to get the array element at position i, then get the first char

When i is 0, (*newval)[0] and *newval[0] happen to yield the same result, so this isn't a functional bug.

However, in the GUC context, newval is a point to a string rather than a two-dimension char array, *newval[i] is
meaningless, so +1 for fixing this to improve readability. 

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/







pgsql-hackers by date:

Previous
From: Ashutosh Sharma
Date:
Subject: Re: synchronized_standby_slots behavior inconsistent with quorum-based synchronous replication
Next
From: Jakub Wartak
Date:
Subject: Re: Add errdetail() with PID and UID about source of termination signal