SourceGraph
SourceGraph copied to clipboard
Apparent inifite time to process code
I using SourceGraph (actually the @utdemir fork, so it'll build) some parser code that I've written to try to identify recursive loops in the type graph and call graph. I seems that I'm managing to throw SourceGraph into an infinite loop, as processing never completes (24 hours later, still going, no files written to SourceGraph/graphs apart from the legend PNGS and a zero length code.dot).
I think it's getting lost in code associated with finding cycles and cliques. Profiling shows this:
COST CENTRE MODULE %time %alloc
pathTree.subPathTree Data.Graph.Analysis.Algorithms.Common 51.9 53.5
pathTree.g' Data.Graph.Analysis.Algorithms.Common 32.7 36.4
pathTree Data.Graph.Analysis.Algorithms.Common 4.0 5.0
makeLeaf.p' Data.Graph.Analysis.Algorithms.Common 3.1 3.1
pathTree.sucs Data.Graph.Analysis.Algorithms.Common 3.1 0.4
I've attached a full profile from the first five minutes of running. SourceGraphProf.zip
Any suggestions on how I can get around this problem?
Yeah, ummm.... I've forgotten to keep this maintained...
It's quite likely that changes I've made to FGL have been reflected here. Any chance of a minimal code sample I can use to try and work out what's going on? And do you know which version of FGL this was built against?
On 8 Sep 2016, at 05:49, Ivan Lazar Miljenovic [email protected] wrote: Any chance of a minimal code sample I can use to try and work out what's going on?
I’m trying to extract an example. Honestly, It’s about 10k lines of code (750 functions and similar data types + constructors) and I’m trying to work out whether a run is going to complete, but just take a while, or just never return is causing difficultly. Unfortunately the code base isn’t public.
I’ll keep trying because I realise this is probably almost impossible without being able to reproduce.
And do you know which version of FGL this was built against?
I’ve built against fgl-5.5.2.3.