Oliver Kruse

I did some tests from 25 digits to 65 digits in steps of 5 digits. It started occuring at 45 digits (when testing 1e5 numbers each). Is a C45 reasonable...

Here is a list of composites that will get stuck in `fmpz_factor` (some of them with small factors; they will eventually get stuck in the QS step): ``` 1000000000000000000000000000000000000000035422 1000000000000000000000000000000000000000042223...

> Can you generate these at various other bit sizes as well? A collection of composites in different sizes: ``` 126582278481012658227848101265822806309 322580645161290322580645161290322654733 555555555555555555555555555555555693427 561482313307130825379000561482313307499 720461095100864553314121037463976947183 1628664495114006514657980456026058632341 1650516364044491319109183307997907145251 1666666666666666666666666666666666845161 1666666666666666666666666666666666979801 1672240802675585284280936454849498335537...

Something I saw in `qsieve_next_A`: First, some polynomials are found when `curr_subset[0] == 0`. After it gets increased, it does not find anything. This again repeats after each factor base...

It is possible that there gets something corrupted. When executing the example in OP with 1000000000000000000000000000000000000000420217, within I minute I got: ```gdb munmap_chunk(): invalid pointer Program received signal SIGABRT, Aborted....

Regarding 1000000000000000000000000000000000000000420217, I found out that `s` will equal `h` at some point in `compute_poly_data.c:273`, which leads to accessing index -1 of an array. More values: ``` (gdb) p s...