ocamlrun should check compatibility of stublibs in CAML_LD_LIBRARY_PATH #6532
Comments
|
Comment author: @gasche We get the following error report¹ from p4bl0, by just running "utop" from a shell in a half-badly-configured OPAM state. *** Error in `/home/p4bl0/.opam/4.02.3/bin/ocamlrun': double free or corruption (out): 0x0000000001314880 *** Edwin thinks is an instance of the present issue. It would be really nice if we could improve the user experience there. |
|
Comment author: @xavierleroy I agree that versioning of DLLs would be a nice touch, but I have no idea how to do it portably. Suggestions welcome. In the meantime, I hope OPAM no longer mucks with CAML_LD_LIBRARY_PATH, as this is generally a recipe for disaster. |
|
Comment author: @xavierleroy I see OPAM still dfeines CAML_LD_LIBRARY_PATH. It must have its reasons. I'm pushing this to after 4.03, as I don't foresee a technical solution in time for 4.03. |
|
Comment author: @gasche Just today another beginner reported the same issue on the #ocaml IRC channel. |
|
Comment author: @avsm OPAM historically sets CAML_LD_LIBRARY_PATH to account for stub libraries installed by a user into the ~/.opam space but not added to the system compiler's ld.conf. Is there a better way to add the ~/.opam stublibs path to the compiler search path other than setting CAML_LD_LIBRARY_PATH? |
|
This issue has been open one year with no activity. Consequently, it is being marked with the "stale" label. What this means is that the issue will be automatically closed in 30 days unless more comments are added or the "stale" label is removed. Comments that provide new information on the issue are especially welcome: is it still reproducible? did it appear in other contexts? how critical is it? etc. |
We're not talking about the system compiler's ld.conf ! ocamlrun from an OPAM switch looks into Hence, setting the
Yes: do not set |
|
@xavierleroy I suspect what you mention doesn't work with system switches where the System switches have been an endless source of problems (I think they should have been killed) but unfortunately it seems users do like them. |
|
The versioning issue here will be dealt with as part of the relocatable compiler (stub libraries get the same treatment as This report pre-dates opam 2.0 which could improve on this very easily by moving the However, that only reduces the problem - I still don't think there's a way to add a directory to the stublib search path when you can't write to |
|
Thanks for the feedback. I concede that CAML_LD_LIBRARY_PATH could need to be set for system switches. But for other switches (the ones I care about) I still contend it should not be set. |
|
Patch set which addresses this is in progress… |
|
PRs for this are actually due soon (there’s a talk on the whole relocatable compiler at this year’s OCaml Workshop) |
Original bug ID: 6532
Reporter: @edwintorok
Status: acknowledged (set by @damiendoligez on 2014-09-15T13:19:56Z)
Resolution: open
Priority: normal
Severity: feature
Platform: x86_64
OS: Linux
OS Version: Linux debian 3.1
Version: 4.02.0+beta1 / +rc1
Target version: later
Category: runtime system and C interface
Monitored by: @avsm
Bug description
My system OCaml is 4.01.0, and for some reason opam had this in variables.sh
(don't know why, running opam config setup -a again fixed it):
CAML_LD_LIBRARY_PATH=/home/edwin/.opam/4.02.0/lib/stublibs:/usr/lib/ocaml/stublibs
This caused utop and utop-full to hang/crash, because ocamlrun 4.02.0 loads dllunix.so (and other stublibs) from CAML_LD_LIBRARY_PATH even if that .so is meant for another OCaml version (4.01.0 in this case).
It would be nice to have a version check to detect mismatches, otherwise applications can crash if CAML_LD_LIBRARY_PATH is set wrongly.
Steps to reproduce
*** Error in `/home/edwin/.opam/4.02.0/bin/ocamlrun': malloc(): memory corruption (fast): 0x0000000000be35d0 ***
Aborted
Additional information
strace /home/edwin/.opam/4.02.0/bin/ocamlrun utop-full 2>&1|grep open|grep dllunix.so
open("/usr/lib/ocaml/stublibs/dllunix.so", O_RDONLY|O_CLOEXEC) = 4
valgrind
which ocamlrunutop-full==7716== Invalid write of size 1
==7716== at 0x4A0BB1B: memcpy@@GLIBC_2.14 (mc_replace_strmem.c:882)
==7716== by 0x5E6F3EE: lwt_unix_blit_string_bytes (in /home/edwin/.opam/4.02.0/lib/stublibs/dlllwt-unix_stubs.so)
==7716== by 0x41C020: caml_interprete (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41D842: caml_main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41A9C0: main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== Address 0x5448ab3 is 531 bytes inside a block of size 1,136 free'd
==7716== at 0x4A08730: free (vg_replace_malloc.c:468)
==7716== by 0x40B289: caml_stat_free (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x413952: caml_input_val (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x4139FE: caml_input_value (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41BE95: caml_interprete (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41D842: caml_main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41A9C0: main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== Thread 2:
==7716== Syscall param read(buf) points to unaddressable byte(s)
==7716== at 0x31E480EADD: ??? (syscall-template.S:81)
==7716== by 0x5E6AC93: worker_bytes_read (in /home/edwin/.opam/4.02.0/lib/stublibs/dlllwt-unix_stubs.so)
==7716== by 0x5E6FEF1: execute_job (in /home/edwin/.opam/4.02.0/lib/stublibs/dlllwt-unix_stubs.so)
==7716== by 0x5E703CA: worker_loop (in /home/edwin/.opam/4.02.0/lib/stublibs/dlllwt-unix_stubs.so)
==7716== by 0x31E48080A3: start_thread (pthread_create.c:309)
==7716== Address 0x692e1d0 is 0 bytes after a block of size 0 alloc'd
==7716== at 0x4A074C0: malloc (vg_replace_malloc.c:291)
==7716== by 0x5C5FF7B: caml_ba_alloc (in /usr/lib/ocaml/stublibs/dllbigarray.so)
==7716== by 0x5C6011D: caml_ba_create (in /usr/lib/ocaml/stublibs/dllbigarray.so)
==7716== by 0x41BF52: caml_interprete (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41D842: caml_main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
==7716== by 0x41A9C0: main (in /home/edwin/.opam/4.02.0/bin/ocamlrun)
The text was updated successfully, but these errors were encountered: