From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jimmy Guo To: Andrew Cagney Cc: Michael Snyder , John David Anglin , gdb-patches@sourceware.cygnus.com Subject: Re: Initialization of hpux_threads Date: Tue, 21 Mar 2000 15:09:00 -0000 Message-id: References: <38D7F9C4.ECF7E17F@cygnus.com> X-SW-Source: 2000-03/msg00446.html I was looking at an old source tree and didn't see its 'revival'. Yes it is some setup to solve the kinds of problem you mentioned. But any other usage of it will only introduce multiple calls to a _initialize_* routine. The initializer in hpux-thread.c is just one of these cases. They're mostly fixed as of today's tree except for a remaining one - remote-nrom.c. - Jimmy Guo, guo@cup.hp.com On Wed, 22 Mar 2000, Andrew Cagney wrote: >> This looks fine. Just an observation: >> CONFIG_OBS is part of DEPFILES, which is part of COMMON_OBS, which is >> part of OBS, which is part of INITFILES. It looks like all three usages >> of CONFIG_INITS in configure.in can be removed. Maybe you can just >> remove CONFIG_INITS altogether in configure.in and Makefile.in. Since >> OBS file list is already looked at by init.c rule, it really makes no >> sense to maintain a CONFIG_INITS file list. > >No. CONFIG_INITS was added because things get complicated when adding >sub-directory files (eg MI) that contain _initialize_*() functions. The >mistake was to assume that it was needed in all (rather than special) >cases. > >By doing this the MI avoided the need to add any nasty initialization >hooks to main() - unlike tui, MPW, and GDBTK. > > Andrew