[master] 8ea006eef does the umem backend affect the amount of malloc NULL returns in vtest?

[Dridi Boukelmoune]

On Mon, Jan 13, 2020 at 5:36 PM Nils Goroll <nils.goroll at uplex.de> wrote:
>
>
> commit 8ea006eef403f2126f399d531568a49d1dafa28b
> Author: Nils Goroll <nils.goroll at uplex.de>
> Date:   Mon Jan 13 18:34:49 2020 +0100
>
>     does the umem backend affect the amount of malloc NULL returns in vtest?

We see segmentation faults on vtest, and since we either assert that
allocations succeed or handle errors, this symptom would be
surprising.

Please note that it did happen even after this patch:

https://varnish-cache.org/vtest/attachment_c186423b5edde99d7fc487cd4b785a7ee73dfdd0_SunOS_i86pc_5.11_1.04_32_sun12.4_uplex_c00057.txt

Dridi

> diff --git a/bin/varnishd/storage/storage_umem.c b/bin/varnishd/storage/storage_umem.c
> index 5ce3ce3d7..644aab5cb 100644
> --- a/bin/varnishd/storage/storage_umem.c
> +++ b/bin/varnishd/storage/storage_umem.c
> @@ -94,7 +94,7 @@ static umem_cache_destroy_f umem_cache_destroyf = NULL;
>  static umem_cache_alloc_f umem_cache_allocf = NULL;
>  static umem_cache_free_f umem_cache_freef = NULL;
>
> -static const char * const def_umem_options = "perthread_cache=0,backend=mmap";
> +static const char * const def_umem_options = "perthread_cache=0";
>  static const char * const env_umem_options = "UMEM_OPTIONS";
>
>  /* init required per cache get:
> _______________________________________________
> varnish-commit mailing list
> varnish-commit at varnish-cache.org
> https://www.varnish-cache.org/lists/mailman/listinfo/varnish-commit