max_cliques() on a directed graph segfaults (C stack overflow) if any_multiple() was called first (igraph 2.3.3)

[jhollway]

What happens, and what did you expect instead?

Calling max_cliques() on a directed graph crashes R with Error: segfault from C stack overflow — but only if any_multiple() has been called on the graph beforehand. Either call alone is fine.

Additional observations:

• max_cliques(g) alone works (returns 12 cliques, with the expected "Edge directions are ignored" warning). any_multiple(g) alone works.
• Wrapping the call in suppressWarnings() does not prevent the crash.
• The crash is specific to any_multiple() as the priming call: which_multiple(g), any_loop(g), or is_connected(g) before max_cliques(g) do not crash.
• Undirected graphs are unaffected: max_cliques(as_undirected(g, mode = "collapse")) after any_multiple() is fine, suggesting the trigger is the directed-graph warning path in max_cliques().
• In a sequence that doesn't crash, the "Edge directions are ignored for maximal clique calculations" warning is mis-attributed to has_multiple_impl (In has_multiple_impl(graph = graph) : Edge directions are ignored..., source cliques/maximal_cliques_template.h:221), which may point to stale condition-handler state left by any_multiple().

Found while testing the manynet package, but reproduces with igraph alone in a fresh session.

To reproduce

library(igraph) # 2.3.3
set.seed(1)
m <- matrix(rbinom(64, 1, .5), 8, 8); diag(m) <- 0
g <- graph_from_adjacency_matrix(m) # directed
invisible(any_multiple(g))
max_cliques(g)
#> Error: segfault from C stack overflow

System information

No response