This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Patch: Enable tree-browser with --enable-checking

Ranjit Mathew wrote:

Bryce McKinlay wrote:

This patch removes the --enable-tree-browser configure option, and instead enables the tree browser when --enable-checking=tree is used (this is the default for development builds).


- tree-browser.o adds only a very small size overhead (approx 20k stripped or 0.5% of cc1's total size on i686)
- its a useful debugging feature
- it sucks to have to remember yet another configure option
- it sucks to have to reconfigure gcc when you just want to browse_tree() during a debugging session
- conceptually it isn't much different from debug_tree and other debugging routines that are always enabled
- without enabling it by default, most developers would not build the tree-browser code, thus it can be subject to bit-rot

Hi Bryce,

 I would argue that Tree Browser be enabled at all
times and not just when "--enable-checking=tree". I for
one regularly bootstrap with "--disable-checking", as
otherwise the bootstraps are even more painful on my
relatively puny machine at home.

Can we please have Tree Browser enabled at all times?

We discussed this on IRC the other day (with Diego and others) and the consensus then was to put it under --enable-checking. But I don't really have an objection to enabling it at all times.

Does anyone object to enabling browse_tree() in all builds?

Perhaps as a compromise we could put the --enable-tree-browser option back in, so that it gets enabled with either --enable-tree-browser or --enable-checking.



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]