Subj : Re: US White House urges devs to dump C and C++
To   : Digital Man
From : tenser
Date : Tue Mar 05 2024 05:47:41

On 01 Mar 2024 at 10:56a, Digital Man pondered and said...
 
 DM>   Re: Re: US White House urges devs to dump C and C++
 DM>   By: tenser to Nightfox on Sat Mar 02 2024 02:30 am
 DM> 
 DM>  > code that appears to work for _years_ can suddenly stop
 DM>  > doing so when you update a point revision of your compiler.
 DM>  > It's madness.
 DM> 
 DM> It's also the reason so many commercial and open source "linters" (static
 DM> analyzers) and run-time analysis tools exist today (primarily, for use
 DM> with C and C++ projects) - detecting undefined behavior, security
 DM> issues. The unsafety of C and C++ have kept a lot of people busy for a
 DM> lot of years. :-)

Oh yeah.  A lot of the argument for C/C++ is that one
can use static analysis tools to catch errors that the
compiler won't; that's all well and good, but yeah, it
keeps us busy.

Also, some bugs are just hard to catch.  At one point,
we got told at Google to stop using unsigned integers
to index into arrays and instead use signed integers,
because unsigned overflow is well-defined in C, but
signed integer overflow is UB, so could be detected with
UBsan.  /shrug

 DM> I've definitely seen what you're describing though - hey this worked
 DM> before I upgraded the tools, it must be a bug in the tools! <sigh> It's
 DM> almost never a bug in the tools. :-(

Yup.  :/

--- Mystic BBS v1.12 A48 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.