Compilation crashes on Slackware Linux, see build output in attachment.
This tracker is for confirmed ATLAS bugs only. Please follow the directions given here:
and start all support requests as items on the support tracker. I will escalate the problem to the bug tracker once it is a confirmed ATLAS bug, and I see it isn't duplicative of a known bug.
That is a strange error. Can you confirm it persists in 3.9.55?
The compilation still crashes. See output of make inside attachment.
OK, something definitely not OK: ATLAS is able to use threads for detecting the number of cores (6), but somehow doesn't generate the needed files for later building.
I wonder if some power-of-2 bigotry has crept back into my code. Just as a WAG, can you try a configure where you add "-t 4" to your configure line and see if things go any better?
I currently don't have any non-power-of-2 machines, and its been a year or two at least since I did.
-t 4 looks better, but there is still some problem.
#I currently don't have any non-power-of-2 machines,
Most BIOSes could disable cores. So when you have e.g. 8 cores, you could easily scale down to 7 cores.
Are you perchance using a heavily loaded shared file system for this install? I just noticed this line right after the "missing" file is supposedly generated:
Timeout waiting for file /usr/src/ATLAS-3.9.55/BUILD/include/atlas_pthreads.h!!
If you are using a shared filesystem, just rule out this problem, how about attempting to install in a local area (eg., /tmp) to rule that out? I think ATLAS has a mechanism for working with very slow filesystems, but it is not guaranteed, and I'd like to know that is the only problem before we try to use it.
Nope. I has been using standalone Linux system. The last time I even did not boot X, I only booted to console to compile ATLAS. May be that I has been running Firefox or Pegasus Mail at same time during compilation of ATLAS.
Just for clearing, I has been running make twice to strip out many succesfull tests. The first run failed and a second run failed on a same place.
OK, can you try 3.9.57?
In the course of readying it, I screwed up configure, which in turn made my install generate the same problem you discuss. This led me to fix a dependence, which *might* fix things for you.
The reason I'm not sure is I don't know why you were hitting that particular dependence anyway: I only hit it because of the error I introduced in archinfo, so most likely the fact that you are hitting it means we've got problems happening earlier . . .
BTW, no need to change # of processors: stuff works fine on my 12-processor system . . .
Full output of make in 3.95.7
Output of configure script 3.9.57
Something has been changed, but compilation still fails. See attached files.
Anyway, the compilation of 3.9.65 runs even much more worse.
I guess that configure script should run smoothly without any crashing as default. If something gets wrong, it should notify me.
Full output from 3.9.65
OK, I don't see any errors in the output file you posted. Did it fail to create a Make.inc or something?
All those warnings coming from gcc are warning my about using the ANSI C standard; they are a bug in gcc which they refuse to fix. I cannot fix them.
Log in to post a comment.
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.