1305396 - mingw-w64 based on GCC >= 5.4.0 fails with "error: 'memmove' was not declared in this scope"

Status: RESOLVED FIXED

(Core :: Security: PSM, defect, P1)

Description

[Georg Koppen]

I finally got a cross-compiler based on GCC 6.2.0 built. But that one breaks compiling Firefox (at least esr45) with

/home/ubuntu/build/tor-browser/security/pkix/lib/pkixnames.cpp: In function 'bool mozilla::pkix::{anonymous}::FinishIPv6Address(uint8_t (&)[16], int, int)':
/home/ubuntu/build/tor-browser/security/pkix/lib/pkixnames.cpp:1701:32: error: 'memmove' was not declared in this scope
           componentsToMove * 2u);
                                ^
make[5]: *** [pkixnames.o] Error 1

Comment 1

[Georg Koppen]

Jacek, back in bug 1199624 you fixed the issues with memcmp and memset. Were there any particular reasons to omit memmove?

Comment 2

[Jacek Caban]

I'm sorry for the delay. I don't remember details of this bug, but I'm pretty sure I simply had no reason to omit memmove as it didn't cause any problem.

I can't really test it now, because tip is badly broken for mingw by configure scripts rewrite :/

Comment 3

[Georg Koppen]

I can actually hit this bug with GCC 5.4.0 already.

Comment 4

[Georg Koppen]

It seems this got fixed between ESR 45 and ESR 52. I thought about finding out what changeset did cause this, but unfortunately I did not have my bisect environment anymore with all the patches needed for dealing with issues between ESR 45 and ESR 52. Setting this up again is not worth the trouble for me. Thus, this is just a simple WORKSFORME. :(

Comment 5

[Georg Koppen]

(In reply to Georg Koppen from comment #4)
> It seems this got fixed between ESR 45 and ESR 52. I thought about finding
> out what changeset did cause this, but unfortunately I did not have my
> bisect environment anymore with all the patches needed for dealing with
> issues between ESR 45 and ESR 52. Setting this up again is not worth the
> trouble for me. Thus, this is just a simple WORKSFORME. :(

Seems I've been too fast. We are hitting this (again) when trying to build for 64-bit Windows (https://trac.torproject.org/projects/tor/ticket/23442).

Comment 6

[Nicolas Vigier]

Attachment: Replace memmove with std::copy_backward in a file that doesn't include cstring explicitly

Here is a proposed patch for fixing that, similar to the fix for bug 1199624.

Comment 7

[Nicolas Vigier]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=404b2e16837515fb83fbf1972542cc1d06c3245b
https://treeherder.mozilla.org/#/jobs?repo=try&revision=d46cb1643533b70bcd2c72632e6d957fe0625dac

Comment 8

[Dana Keeler (she/her) (use needinfo) [:keeler]]

Comment on attachment 8907854 [details] [diff] [review]
Replace memmove with std::copy_backward in a file that doesn't include cstring explicitly

Review of attachment 8907854 [details] [diff] [review]:
-----------------------------------------------------------------

Just a heads-up: Mozilla style is to include 8 lines of context in patches.
This looks good, but I have a simplifying change I think we should make. r- for now.

::: security/pkix/lib/pkixnames.cpp
@@ +1707,4 @@
>    // Shift components that occur after the contraction over.
>    size_t componentsToMove = static_cast<size_t>(numComponents -
>                                                  contractionIndex);
> +  std::copy_backward(address + (2u * static_cast<size_t>(contractionIndex)),

Looks like this needs a #include <algorithm> in this file (the style is to list it before other #includes, with a blank line after it).

@@ +1707,5 @@
>    // Shift components that occur after the contraction over.
>    size_t componentsToMove = static_cast<size_t>(numComponents -
>                                                  contractionIndex);
> +  std::copy_backward(address + (2u * static_cast<size_t>(contractionIndex)),
> +                     address + (2u * static_cast<size_t>(contractionIndex))

Actually, I think we can simplify this by getting rid of componentsToMove altogether and just using `address + (2u * static_cast<size_t>(numComponents))` here.

Comment 9

[Nicolas Vigier]

Attachment: Replace memmove with std::copy_backward in a file that doesn't include cstring explicitly

Comment 10

[Nicolas Vigier]

Thanks for the review and sorry for taking a long time to reply. I attached a new patch taking into account your comments.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=ad8684558b50382d90beaa35b420fe3af878c6ab

Comment 11

[Dana Keeler (she/her) (use needinfo) [:keeler]]

Comment on attachment 8919135 [details] [diff] [review]
Replace memmove with std::copy_backward in a file that doesn't include cstring explicitly

Review of attachment 8919135 [details] [diff] [review]:
-----------------------------------------------------------------

Great - thanks.

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/0a5c3789fac1
Replace memmove with std::copy_backward in a file that doesn't include cstring explicitly. r=keeler

Comment 13

[Stefan Hindli [:stefan_hindli]]

https://hg.mozilla.org/mozilla-central/rev/0a5c3789fac1