On Mon, Jan 5, 2015 at 5:12 AM, Andrew Haley <aph@xxxxxxxxxx> wrote: > On 05/01/15 10:29, Cyd Haselton wrote: >> On Mon, Jan 5, 2015 at 3:52 AM, Andrew Haley <aph@xxxxxxxxxx> wrote: >>> On 05/01/15 01:23, Cyd Haselton wrote: >>> >>> Hmm. Right now I would have thought that the only command you'd need >>> to run on both was the one which failed on 4.9, whatever that was. >> >> See the list of commands I posted for the 4.8.4 libgcc build...both >> are exactly the same. > > Okay, so the input files are different. Look at whichever input file > supplies dlopen in 4.8 and see if it doesn't supply dlopen in 4.9. > ...okay. (There'sma reason why I use ''*.o' as a placeholder for the input files...it is an obnoxiously long list. I'll see what I can do) >> I can try again to manually re-run both commands (see previous emails >> for exact command) but if one succeeds and one fails i'm not sure how >> that would help diagnose the issue or provide info that would do the >> same. > > It gives us exactly what we need. Look in the input files. See > which one supplies dlopen. > >> Wouldn't it be more helpful if I post the differences between both >> buuld dirs...something that would indicate why one build failed and >> the other didn't? > > You need to look inside the files with "nm" or something similar. > Will do. Right now i'm re-running make in the 4.9 libgcc build dir, after having run make clean && make in the 4.8 libgcc build dir. I've tee-ed both and will look at results. > Andrew. > >
