On Tue, Oct 27, 2026 at 4:53 AM Ales Musil <[email protected]> wrote:
>
>
>
> On Mon, Oct 19, 2026 at 10:53 PM Mark Michelson via dev 
> <[email protected]> wrote:
>>
>> This is similar to vector_clone(), except that it performs a shallow
>> copy of the vector instead of using memcpy(). The vector that is stolen
>> from is for all intents and purposes unusable after calling this
>> function.
>>
>> This is useful for cases where a vector is built up and then needs to be
>> copied as part of a struct initializer.
>>
>> Signed-off-by: Mark Michelson <[email protected]>
>> ---
>> v1 -> v2:
>>  * Expanded the test case in test-vector.c.
>> ---
>>  lib/vec.c           | 18 +++++++++++++++
>>  lib/vec.h           |  1 +
>>  tests/ovn.at        |  2 ++
>>  tests/test-vector.c | 54 +++++++++++++++++++++++++++++++++++++++++++++
>>  4 files changed, 75 insertions(+)
>>
>> diff --git a/lib/vec.c b/lib/vec.c
>> index 97e6b6549..e4bb72c2c 100644
>> --- a/lib/vec.c
>> +++ b/lib/vec.c
>> @@ -168,6 +168,24 @@ vector_clone(struct vector *vec)
>>      return clone;
>>  }
>>
>> +/* Performs a shallow copy of vec into another structure. vec is
>> + * re-initialized after this operation and is safe to re-use or
>> + * destroy. However, its buffer will be empty.
>> + */
>> +struct vector
>> +vector_steal(struct vector *vec)
>> +{
>> +    struct vector thief = (struct vector) {
>> +        .buffer = vec->buffer,
>> +        .esize = vec->esize,
>> +        .len = vec->len,
>> +        .capacity = vec->capacity,
>> +    };
>> +
>> +    vector_init(vec);
>> +    return thief;
>> +}
>> +
>>  /* Reserves additional space to fit extra n items. */
>>  void
>>  vector_reserve(struct vector *vec, size_t n)
>> diff --git a/lib/vec.h b/lib/vec.h
>> index ff8f69a5e..50dd09bc4 100644
>> --- a/lib/vec.h
>> +++ b/lib/vec.h
>> @@ -76,6 +76,7 @@ bool vector_remove_block(struct vector *vec, size_t start, 
>> size_t end);
>>  void *vector_get_ptr(const struct vector *vec, size_t index);
>>  void vector_shrink_to_fit(struct vector *vec);
>>  struct vector vector_clone(struct vector *vec);
>> +struct vector vector_steal(struct vector *vec);
>>  void vector_reserve(struct vector *vec, size_t n);
>>
>>  /* Pushes element into the end of the vector. */
>> diff --git a/tests/ovn.at b/tests/ovn.at
>> index 94ed06c63..8081fb04c 100644
>> --- a/tests/ovn.at
>> +++ b/tests/ovn.at
>> @@ -2453,6 +2453,8 @@ check ovstest test-vector shrink
>>  check ovstest test-vector clone
>>
>>  check ovstest test-vector pointers
>> +
>> +check ovstest test-vector steal
>>  AT_CLEANUP
>>
>>  AT_SETUP([Sparse array operations])
>> diff --git a/tests/test-vector.c b/tests/test-vector.c
>> index f2bad0c33..f91741856 100644
>> --- a/tests/test-vector.c
>> +++ b/tests/test-vector.c
>> @@ -254,6 +254,59 @@ test_clone(struct ovs_cmdl_context *ctx OVS_UNUSED)
>>      vector_destroy(&clone);
>>  }
>>
>> +static void
>> +test_steal(struct ovs_cmdl_context *ctx OVS_UNUSED)
>> +{
>> +    uint32_t elements[3] = {0, 1, 2};
>> +    struct vector vec = VECTOR_CAPACITY_INITIALIZER(uint32_t, 10);
>> +    vector_push_array(&vec, elements, ARRAY_SIZE(elements));
>> +
>> +    struct vector thief = vector_steal(&vec);
>> +    ovs_assert(vector_capacity(&thief) == 10);
>> +    ovs_assert(vector_len(&thief) == 3);
>> +
>> +    ovs_assert(vector_capacity(&vec) == 0);
>> +    ovs_assert(vector_len(&vec) == 0);
>> +
>> +    size_t i = 0;
>> +    uint32_t num;
>> +    VECTOR_FOR_EACH (&thief, num) {
>> +        ovs_assert(elements[i++] == num);
>> +    }
>> +    ovs_assert(i == ARRAY_SIZE(elements));
>> +
>> +    vector_shrink_to_fit(&thief);
>> +    ovs_assert(vector_capacity(&thief) == 3);
>> +    ovs_assert(vector_len(&thief) == 3);
>> +
>> +    /* Let's make sure the original vec has been unaffected by
>> +     * the changes to thief.
>> +     */
>> +    ovs_assert(vector_capacity(&vec) == 0);
>> +    ovs_assert(vector_len(&vec) == 0);
>> +
>> +    /* The vast majority of the time, the reason you would want to steal a
>> +     * vector is because it is going to go out of scope and this prevents
>> +     * having to destroy it. However, it is also possible to steal a vector 
>> as
>> +     * a way of saving its contents elsewhere and resetting it. Let's make 
>> sure
>> +     * we can still operate on vec after it has been stolen.
>> +     */
>> +    uint32_t more_elements[5] = {3, 4, 5, 6, 7};
>> +    vector_push_array(&vec, more_elements, ARRAY_SIZE(more_elements));
>> +    ovs_assert(vector_capacity(&vec) == 5);
>> +    ovs_assert(vector_len(&vec) == 5);
>> +
>> +    /* And now let's make sure the thief is unaffected by changes to vec. */
>> +    ovs_assert(vector_capacity(&thief) == 3);
>> +    ovs_assert(vector_len(&thief) == 3);
>> +
>> +    /* Since we have re-added elements to vec, we now need to destroy it.
>> +     * If we had stolen from it and left it alone, this would be 
>> unnecessary.
>> +     */
>> +    vector_destroy(&vec);
>> +    vector_destroy(&thief);
>> +}
>> +
>>  static void
>>  test_pointers(struct ovs_cmdl_context *ctx OVS_UNUSED)
>>  {
>> @@ -305,6 +358,7 @@ test_vector_main(int argc OVS_UNUSED, char *argv[] 
>> OVS_UNUSED)
>>          {"shrink", NULL, 0, 0, test_shrink, OVS_RO},
>>          {"clone", NULL, 0, 0, test_clone, OVS_RO},
>>          {"pointers", NULL, 0, 0, test_pointers, OVS_RO},
>> +        {"steal", NULL, 0, 0, test_steal, OVS_RO},
>>          {NULL, NULL, 0, 0, NULL, OVS_RO},
>>      };
>>      struct ovs_cmdl_context ctx;
>> --
>> 2.55.0
>>
>> _______________________________________________
>> dev mailing list
>> [email protected]
>> https://mail.openvswitch.org/mailman/listinfo/ovs-dev
>>
>
> Looks good to me, thanks.
> Acked-by: Ales Musil <[email protected]>
>
> I have also stopped the review at this patch, it seems that the rest of the
> series has some issue with CI, could you please check it out?
>

Thanks Ales. The problem with CI was because in patch 11, I updated
the includes in en-route-policies.h to have

#include <arpa/inet.h>

On clang, this works fine. But the gcc CI lanes complain because I
need to also have

#include <netinet/in.h>

before. I have corrected this and have a clean CI run here:
https://github.com/putnopvut/ovn/actions/runs/41495678921

I'll submit a v3 that has the preliminary patch I suggested in my
reply to your review of patch 2, as well as the #include fix for patch
11.

>

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to