Re: [PATCH v2] mceliece6688128f: fix stack overflow crash on win64/wine

Jacob Bachmeyer via Gcrypt-devel <[email protected]> Sun, 28 Sep 2025 18:23:20 -0500
Newsgroups gmane.comp.encryption.gpg.libgcrypt.devel
Message-ID <[email protected]>
On 9/28/25 02:56, Jussi Kivilinna wrote:
> On 28/09/2025 04:33, Jacob Bachmeyer via Gcrypt-devel wrote:
>> [...]
>>
>> Would this make sense as "const unsigned int sizeof_mat" to enable 
>> better compiler optimizations/inlining since PK_NROWS and nblocks_H 
>> seem to be constants?  Or is GCC able to deduce that sizeof_mat is 
>> only assigned once?
>
> With GCC -O0, "unsigned int sizeof_mat" and "const unsigned int 
> sizeof_mat" generate the same code, value is assigned to stack memory. 
> With -O1 and above, compiler deduces that this is constant variable 
> and optimizes as it sees best fit. 'const' on stack variables only 
> really affects syntax checks and gives compile-time error when trying 
> reassigning to that variable.

So GCC does recognize constants and optimize accordingly.  Good. Why not 
include the "const" as well to ensure that a compile-time error will be 
thrown if any future code attempts to change the value, since there is 
no way that such code could be correct?


-- Jacob



_______________________________________________
Gcrypt-devel mailing list
[email protected]
https://lists.gnupg.org/mailman/listinfo/gcrypt-devel