Dmytro Milinevskyy wrote:

Hello, we've encountered this problem mainly on Android device. The kernel sends a fake signal and the baking storage(flash) is not so fast. I've noticed a typo in my patch for the macro DO_PWRITE in mdb_env_init_meta. The test should be done for EINTR, not EINVAL. Do you want me to resend the patch?

Yes I already noticed that, so I wrote the patch myself instead of applying yours.

Thank you for the context, that's good information.

Thanks, Dmytro

On Fri, Apr 17, 2015 at 7:34 PM, Howard Chu <hyc@symas.com mailto:hyc@symas.com> wrote:

milinevskyy@gmail.com <mailto:milinevskyy@gmail.com> wrote:

    Full_Name: Dmytro Milinevskyy
    Version:
    OS: Linux
    URL: ftp://ftp.openldap.org/incoming/
    Submission from: (NULL) (92.132.131.166)


    On linux systems it's common to have system calls interrupted
    and restarted.
    LMDB should retry pwrites to avoid "false" failures.


In practice, writes to a file are "fast" and will always write at
least one byte. But we've patched this as suggested, thanks.

--
   -- Howard Chu
   CTO, Symas Corp. http://www.symas.com
   Director, Highland Sun http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP http://www.openldap.org/project/