430680 - Backport GCC patch to GCC 4.3

Status: RESOLVED FIXED

(Developer Infrastructure :: Source Code Analysis, defect)

Description

[(dormant account)]

Too many people are getting confused by having to keep up with mainline.

Also should update docs wgeting gcc 4.3 instead of using hg.

Comment 1

[Vlad Sukhoy]

One particular issue I've noticed with the current trunk approach is that the patches tend to bitrot against the actual gcc trunk, so people have to checkout trunk from some earlier date. Often said people are new to mercurial and thus have to lookup the documentation and then come up with a good date. All this isn't very convenient. Using a snapshot would certainly help to improve the situation as more people evaluate xhydra for their needs.

Comment 2

[(dormant account)]

The current situation is an artifact of the hg gcc mirror not having svn stable branches on it and me being lazy. 
I'm sorry for the amount of hassle this causes to people :)

Comment 3

[(dormant account)]

Updated patches and http://developer.mozilla.org/en/docs/Dehydra_GCC

Now have to fix the location testcase.

Comment 4

[David Mandelin [:dmandelin]]

(In reply to comment #3)
> Updated patches and http://developer.mozilla.org/en/docs/Dehydra_GCC
> 
> Now have to fix the location testcase.

Test case is correct, it's actual 4.3 bustage. Problem is that some things that are location_t in the later GCC were 'source_locus' in GCC 4.3, which was a typedef name for 'location_t *'. Example: gimple stmt at tree.h:1844. source_locus is defined at input.h:68. 

When fixed, if you do 'grep locus treehydra_generated.c', every line in the result should call convert_location_t (I think).

I'm gonna bounce it back to you because I figure you want to handle convert_tree.js changes. Just rebounce it if you don't. :-)

Comment 5

[(dormant account)]

pushed, tests pass!