In my work, I run codes under CENTOS 6.9 (jobs environment) and OMV LX 3.03 (home). The same code is compiled in both places with gfortran 7.2. Now, it happens that I’m getting segmentation fault at home where nothing wrong happens at my job.
After investigating the source of segmentation fault I could see that there was nothing wrong at all in the code. In fact, changing the compilation option from debug to some level of optimization generates a code that runs with no problem in OMV LX 3.03.
Does someone have any idea why gfortran at OMV LX 3.03 would be lacking something or some other reason for a problem like this?
I’m sorry not being able to provide a simple testing code. I know this seems to be a rather vague question …
OpenMandriva does not have enough people involved (as contributors) for that level of attention to detail. Also the nature of a Community distro with all volunteer contributors is that contributors tend to spend most of their time on the fun things they enjoy doing not on boring details that would only matter to a handful of elite users. Human nature.
If that sounds like I’m apologizing or trying to alibi I am not. This is the nature of the distro as it exists right now. To change requires an awareness of initial conditions.
Sorry not to have updated this topic with new information.
Yes, there is a difference in gcc-gfortran 7.2 in OMV LX 3.03 and the CENTOS I use in my job (version 7.2 compiled by a colleague, not the distro’s version). However, now I believe the difference makes me prefer the OMV LX 3.03 one.
When i create this topic I believed there were no errors in my code and then something should be wrong with OMV gfortran. That wasn’t true. The code had one (or more) memory violation and execution in CENTOS did not showed any problem. OMV LX 3.03 code crashed with segmentation fault. So OMV LX 3.03 code was “right”, CENTOS code (by the extra gfortran) was misleading.
Now, no compilation flags changes the result with OMV’s gfortran as expected, It is OK now.
Anyway, asap I will look for details on the CENTOS compilation of gfortran 7.2 and I will post here just for the case someone may be curious about the causes for that.
