On Mon, 18 Feb 2019, John Helmert III via curl-library wrote:

Valgrind reports an immense number of errors for me. Running valgrind like `valgrind curl https://example.com`. Tail end of that output:

Thanks for your report.

This is curious. We run valgrind all the time - on buils done on every commit, and several of us do it locally too in our development machines and this is not something we see, neither on the CI builds nor elsewhere! We *also* run fuzzers and sanitizers on the builds and they didn't flag this either.

Built/installed with:

`./buildconf`
`./configure`
`make -j8`
`sudo make install`

In an attempt to reproduce this, here's me on Linux 4.19, glibc 2.28, gcc 8.2.0, valgrind 3.14.0, openssl 1.1.1a on Debian unstable.

I added '--disable-shared' to configure so that I can run valgrind on the built src/curl binary without it being a shell script or me having to run make install first.

I tried this just now and... nothing:

$ valgrind ./src/curl https://example.com
...
==10829==
==10829== HEAP SUMMARY:
==10829==     in use at exit: 0 bytes in 0 blocks
==10829== total heap usage: 32,248 allocs, 32,248 frees, 3,259,854 bytes allocated
==10829==
==10829== All heap blocks were freed -- no leaks are possible
==10829==
==10829== For counts of detected and suppressed errors, rerun with: -v
==10829== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)


Clearly we need something extra to happen!

Can you rerun configure with '--enable-debug' to get debug symbols on the library and then rerun the experiment to get more details on the backtrace for where exactly in the code valgrind finds the problems?

--

 / daniel.haxx.se
-------------------------------------------------------------------
Unsubscribe: https://cool.haxx.se/list/listinfo/curl-library
Etiquette:   https://curl.haxx.se/mail/etiquette.html

Reply via email to