Bug #15520: [patch] configure should refuse to build with jemalloc when headers are missing

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.7.0dev (2019-01-10 fix_jemalloc_i.. 66768) [x86_64-darwin18]
• Backport: 2.4: REQUIRED, 2.5: DONE, 2.6: DONE

[mistydemeo (Misty De Meo)]

When the --with-jemalloc option is passed, the configure script will correctly fail with an error if the library is missing. However, if the library is present and headers are missing, configure will succeed and allow the build to proceed. The actual build is unlikely to succeed in this case since the functions won't be defined.

Steps to reproduce:

1. Install jemalloc.
2. Delete jemalloc's headers.
3. Run ./configure --with-jemalloc &&is make

Expected behaviour:
configure fails because the requested jemalloc's headers are missing

Actual behaviour:
configure succeeds. On macOS, the build then fails with the following:

compiling gc.c
gc.c:7955:12: error: implicit declaration of function 'malloc_usable_size' is invalid in C99
      [-Werror,-Wimplicit-function-declaration]
    return malloc_usable_size(ptr);
           ^
gc.c:7955:12: note: did you mean 'malloc_good_size'?
/Library/Developer/CommandLineTools/SDKs/MacOSX10.14.sdk/usr/include/malloc/malloc.h:134:15: note:
      'malloc_good_size' declared here
extern size_t malloc_good_size(size_t size);
              ^
1 error generated.
make: *** [gc.o] Error 1

The attached patch fixes this by ensuring that configure fails if --with-jemalloc is passed but the headers are missing in the same way that it will fail if the library is missing.

This patch is against trunk. I've also reproduced it in several previous versions of Ruby.

[shevegen (Robert A. Heiler)]

The actual build is unlikely to succeed in this case since the functions won't be defined.

Agreed, makes sense to me. I remember that I may run into similar situations every now
and then when I e. g. wish to upgrade some program manually on my own but may have
forgotten to remove all old parts of that program (e. g. under the /usr/ hierarchy). I don't
know if this applies to Mac but on Linux I remember similar situations here and there.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

Applied in changeset trunk|r66779.

---

configure: refuse to build with jemalloc when header is missing

[ruby-core:90964] [Bug #15520]

Freom: Misty De Meo [e-mail]

[mistydemeo (Misty De Meo)]

Thanks for the quick response! Could you backport this to the active backport branches, too?

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.4: UNKNOWN, 2.5: UNKNOWN, 2.6: UNKNOWN to 2.4: REQUIRED, 2.5: REQUIRED, 2.6: REQUIRED

[naruse (Yui NARUSE)]

• Backport changed from 2.4: REQUIRED, 2.5: REQUIRED, 2.6: REQUIRED to 2.4: REQUIRED, 2.5: REQUIRED, 2.6: DONE

ruby_2_6 r66852 merged revision(s) 66779.

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.4: REQUIRED, 2.5: REQUIRED, 2.6: DONE to 2.4: REQUIRED, 2.5: DONE, 2.6: DONE

ruby_2_5 r67240 merged revision(s) 66779.