Re: [PATCH] CLEANUP: h2: Help static analyzers understand the list's end marker

2020-03-19 Thread Tim Düsterhus
Willy, Am 19.03.20 um 15:55 schrieb Willy Tarreau: > Actually I'm pretty sure that I did it this way precisely for performance > reasons: avoid repeatedly checking a pointer for half of the headers which > are pseudo headers (method, scheme, authority, path just for the request). > > It's

Re: [PATCH] CLEANUP: h2: Help static analyzers understand the list's end marker

2020-03-19 Thread Willy Tarreau
Hi Tim, On Thu, Mar 19, 2020 at 03:15:24PM +0100, Tim Duesterhus wrote: > Willy, > > I know you dislike adjusting code to please static analyzers, but I'd argue > that using the new IST_NULL + isttest() combination is easier to understand > for humans as well. A simple .ptr == NULL check might

[PATCH] CLEANUP: h2: Help static analyzers understand the list's end marker

2020-03-19 Thread Tim Duesterhus
Willy, I know you dislike adjusting code to please static analyzers, but I'd argue that using the new IST_NULL + isttest() combination is easier to understand for humans as well. A simple .ptr == NULL check might also be slightly faster compared to isteq() with an empty string? I have verified