Re: Fix %Z parsing in strptime [BZ #16088]

[Paul Eggert <eggert@cs.ucla.edu>]

On 2023-07-17 05:57, Wilco Dijkstra via Libc-alpha wrote:
> Hi Stanley,
> 
>> +	  /* we recognize the format [-+a-zA-Z0-9]{3,} */
> 
> Is that really the correct format? Eg. should it be able to parse UTC offsets like UTC+04:30?

strftime doesn't generate such offsets, so this is asking whether 
strptime should accept forms that strftime doesn't generate, at the cost 
of strptime misparsing some unusual forms that strftime does generate 
(e.g., strftime with "%Z:%Y").


> This is basically isalnum() but not allowing '+' or '-'.

Yes, '+' and '-' should be allowed.


> rather than allowing any sequence of these characters, would it
> not be better to scan just for isalpha() first, and only if you see '+' or '-' check the UTC
> offset syntax?

POSIX allows settings like TZ='<-X->0', unfortunately, which generate 
time zone abbreviations like "-X-". Also, an actual isalpha would be 
dicey, as abbreviations use ASCII letters (OK, letters in the POSIX 
locale, basically the same thing), so it's arguably better to avoid 
isalpha here.