Re: Make and makefile bugs (PRs 49085, 49086, 49087)

[Antoine LECA <antoine.leca.1%gmail.com@localhost>]

On 2014-08-18 14:21Z, Jarmo Jaakkola wrote:
> On Mon, Aug 18, 2014 at 03:17:44PM +0200, Antoine LECA wrote:
>> On 2014-08-17 15:03Z, Jarmo Jaakkola wrote:
>>> So, if .a, .b, and .a.b are known suffixes,
>>
>> According to POSIX, .a.b is not a valid suffix:
>>      [...] (where .s1 and .s2 are suffixes that have been given as
>>       prerequisites of the .SUFFIXES special target and s1 and s2 do
>>       not contain any <slash> or <period> characters.)
> 
> Yeah, I know.  But our implementation allows any string not containing
> a slash as an extension, thus we need to deal with this possibility.

The point was, we are already in grey territory (that is, outside of
POSIX compliance); since there is a known-to-be-better solution
(patterns) for the long term, perhaps it would be a better idea to not
loose too much time on awkward cases like .a.b suffixes?

Furthermore, I am not yet convinced (particularly after a quick review
of the code, where I can spot a number of strrchr() calls) that it deals
"correctly" with .a.b (or more likely, .tar.gz) suffixes in any cases of
uses of implicit rules.

Also, I only mentioned <slash> because Posix does (as does my own
parsing code), and it was easier to copy-and-paste :-) Please drop it.


Antoine