./configure failing when --with-memory-sanitizer specified

[JaonHax]

BPO	42268
Nosy	@pablogsal, @erlend-aasland, @JaonHax
PRs	bpo-42268: Fail the configure step if the selected compiler doesn't support memory sanitizer #29806 [3.10] bpo-42268: Fail the configure step if the selected compiler doesn't support memory sanitizer (GH-29806) #29815
Files	configure-out.txt: ./configure failed output configure-out-1.txt: ./configure successful output

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2021-11-28.21:24:58.253>
created_at = <Date 2020-11-05.00:37:06.343>
labels = ['3.10', 'type-bug', '3.9', 'build', '3.11']
title = './configure failing when --with-memory-sanitizer specified'
updated_at = <Date 2021-11-28.21:24:58.252>
user = 'https://github.com/JaonHax'

bugs.python.org fields:

activity = <Date 2021-11-28.21:24:58.252>
actor = 'pablogsal'
assignee = 'none'
closed = True
closed_date = <Date 2021-11-28.21:24:58.253>
closer = 'pablogsal'
components = ['Build']
creation = <Date 2020-11-05.00:37:06.343>
creator = 'JaonHax'
dependencies = []
files = ['49569', '49570']
hgrepos = []
issue_num = 42268
keywords = ['patch']
message_count = 6.0
messages = ['380386', '380389', '380392', '406739', '407159', '407232']
nosy_count = 3.0
nosy_names = ['pablogsal', 'erlendaasland', 'JaonHax']
pr_nums = ['29806', '29815']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue42268'
versions = ['Python 3.9', 'Python 3.10', 'Python 3.11']

[JaonHax]

Exact command run is as follows:
sudo ./configure --with-pymalloc --with-assertions --enable-optimizations --enable-loadable-sqlite-extensions --enable-shared --with-address-sanitizer --with-undefined-behavior-sanitizer --with-memory-sanitizer -C

(ran with root because it seemed to make more things succeed in the checks without those flags)

The command's output is attached in configure-out.txt

[JaonHax]

Additional note, since I forgot it in the initial message: I'm building Python 3.9.0 from source on Debian v10, specifically the Debian VM on Chromebooks (which I believe is called Crostini).

I've made sure to install all the proper dependencies, as far as I know, and I'm 99% sure that I can successfully build it if I remove the --with-memory-sanitizer flag. Running the command again without it to double-check.

[JaonHax]

Yeah, exactly as I thought, ran the exact same command except without --with-memory-sanitizer and the configuration was successful. The successful output has been attached in configure-out-1.txt for cross-comparison, if it's necessary.

[erlend-aasland]

Reproducible on 3.9, 3.10, and on main.

The --with-memory-sanitizer option adds "-fsanitize=memory -fsanitize-memory-track-origins=2 -fno-omit-frame-pointer" to CFLAGS, and "-fsanitize=memory -fsanitize-memory-track-origins=2" to LDFLAGS. The MemorySanitizer is a clang specific feature, AFAIK. You're using GCC, so what happens is that these (GCC incompatible) flags are added and the following compile check (getaddrinfo) fails.

IMO, adding a compiler check and printing a warning if --with-memory-sanitizer is used with GCC should be an ok solution. Alternatively, we inform the user that we're switching to using the leak sanitiser, if we're on GCC.

+pablo

See also:

• https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html
• https://clang.llvm.org/docs/MemorySanitizer.html

[pablogsal]

New changeset e71c12e by Pablo Galindo Salgado in branch 'main':
bpo-42268: Fail the configure step if the selected compiler doesn't support memory sanitizer (GH-29806)
e71c12e