Re: RFA: fix PR middle-end/23135

[Bernd Schmidt <bernds_cb1 at t-online dot de>]

Joern RENNECKE wrote:
> Bernd Schmidt wrote:
>
> >  
> > My brain hurts, but...
> >
> > > 2005-07-27  J"orn Rennecke <joern.rennecke@st.com>
> > >
> > >     * reload.c (find_reloads_subreg_address): Pass down TYPE
> > >     unchanged.  Change all callers except find_reloads_toplev.
> >
> >
> >
> > This seems ok (along with the testcase, although I wouldn't 
> > necessarily include the RTL dump in the comments).
>
>
>
> Thanks for the review.
> If I understand you correctly,  you are not actually opposed to having 
> these comments,
> but wouldn't put them there yourself.  So I'll rather leave them in.  
> Comments cost almost
> nothing in tetsuite run time, and now that the preprocessor is 
> integrated into cc1, they
> also don't cause the inconvenience of having to preprocess the file 
> separately first to
> run cc1 under gdb control.  I have found in the past it can take a day 
> to get the right
> version of sources checked out & built to understand why a particular 
> piece of code is
> needed in reload.  With the reload dump, the rtl of the reloaded insn 
> before reloading,
> and a few explanatory sentences, anyone who knows the reload types can 
> see the issue
> without having to go to such lengths, and those who don't are at least 
> given some clues
> what to look up.

Eh, well, that's fine, really.  I just figure that over time the comment 
would no longer match the RTL generated by the compiler; a URL pointing 
to discussion of the issue might be of more value.  But. really, feel 
free to check in the test case in any form you like.


Bernd