configure template-matching change
От | Tom Lane |
---|---|
Тема | configure template-matching change |
Дата | |
Msg-id | 15452.911797338@sss.pgh.pa.us обсуждение исходный текст |
Ответы |
Re: [HACKERS] configure template-matching change
|
Список | pgsql-hackers |
I've just checked in a rewrite of configure's method for guessing the right template file to use when --with-template is not provided. The old method was to look for an exact match to the $host value (from config.guess) in template/.similar, and if that wasn't found to strip all version information (digits and dots at the end) from $host and try to match that against prefixes of entries in .similar. This is clearly broken, because given .similar entries likesparc-sun-sunos4=sunos4_gccsparc-sun-sunos5=solaris_sparc_gcc it is incapable of correctly choosing the one to use for a $host value like "sparc-sun-sunos4.1.3". After you strip the "4.1.3" there is no basis for choosing the right .similar entry. What I checked in tries for an exact match, and then tries to match .similar entries to prefixes of the $host string. This is fairly obviously the right way round, IMHO. I was able to remove some redundant entries in .similar after making the change; for instancei386-pc-bsdi3.0=bsdi_2.1i386-pc-bsdi3.1=bsdi_2.1 collapse toi386-pc-bsdi3=bsdi_2.1 which has some chance of working on bsdi 3.2 as well, if there is such a thing. *However*, I'm not too sure that the shell-script constructs I used are 100% portable. If you find that configure now goes belly-up when you leave off --with-template, please let me know. I checked this into the 6.4 tree as well, since it is in response to a recent complaint that 6.4 fails to configure correctly on SunOS 4.1.3... regards, tom lane
В списке pgsql-hackers по дате отправления: