On Mon May 07 04:28:22 2007, codermattie wrote: > hello, > > Building on commit #18446 this patch honors PARROT_PREFER_SOURCE. > > quick background: > > to get the install/working-copy to work the same extensions should > not be used with .load_bytecode. Sometimes it will be compiled, > sometimes not. > > The default behavior implemented in #18446 was to first preserve > compatibility with the existing code-base by trying a path as > it was given first. If that fails it will try appending extensions, > from the lowest level compiled objects (.pbc) to the highest level > source > (.pir). > > This is the perl5 use behavior and what you would expect running out > of a install target. > > Some developers did not like having to run a "make clean" to run the > latest version of the source if there were stale objects lying around. > > This patch addresses that issue by honoring PARROT_PREFER_SOURCE. > Currently > the value is ignored. If it is set in the environment variable table > it > will reverse the normal extension guessing order, trying the highest > level form of the code first. Developers who wish to avoid a make > clean > can now either avoid compiling files, or simply set this environment > variable. > > As noted in the comments if the value of PARROT_PREFER_SOURCE is ever > interpreted > I think it should be a path spec for directories where source is > preferred over > compiled objects. Currently that is not implemented, but it can be if > someone > makes a more concrete case for it. > > This patch should be discussed, during which I will add the necessary > tests > and documentation changes. I have deferred this until now , when the > behavior > is complete enough to document. > > patch: > > This patch is based off rev 18446 with the spelling fixes from #42897 > applied. > > harness: > > when I ran the test-suite I got a failure from t/perl/Parrot_Docs.t. A > subsequent > run of prove to double check it resulted in a pass. Strange. End > result the > test-suite passed unless someone can reproduce the failure with the > harness. > > Cheers, > Mike Mattie - codermattie <!-- x --> at gmail.com
There hasn't been any discussion about this patch, and it looks pretty good to me, so I've applied it as r18482, thanks! I haven't noticed the failure from t/perl/Parrot_Docs.t. This might be a problem with your setup as you mentioned in a previous email. Paul
