1037344 - E/GeckoLinker( NNNN): /data/data/com.adobe.flashplayer/lib/libysshared.so: Text relocations are not supported

Status: RESOLVED INCOMPLETE

(Core :: mozglue, defect)

Description

[Ed Morley [:emorley]]

In some passing Android jobs, I'm seeing:
23:36:26     INFO -  07-10 23:36:15.617 E/GeckoLinker( 4421): /data/data/com.adobe.flashplayer/lib/libysshared.so: Text relocations are not supported

eg:
https://tbpl.mozilla.org/php/getParsedLog.php?id=43585842&tree=Mozilla-Inbound
Android 4.0 Panda mozilla-inbound opt test robocop-2 on 2014-07-10 23:13:24 PDT for push e08b171b9a02

https://tbpl.mozilla.org/php/getParsedLog.php?id=43587982&tree=Mozilla-Inbound
Android 4.0 Panda mozilla-inbound opt test robocop-1 on 2014-07-10 23:48:45 PDT for push e08b171b9a02

https://tbpl.mozilla.org/php/getParsedLog.php?id=43585083&tree=Mozilla-Inbound
Android 4.0 Panda mozilla-inbound talos remote-tp4m_nochrome on 2014-07-10 23:03:21 PDT for push 805ac89b5924

This comes from one of:

http://mxr.mozilla.org/mozilla-central/source/mozglue/linker/CustomElf.cpp#582
578       case DT_TEXTREL:
579         if (strcmp("libflashplayer.so", GetName()) == 0) {
580           has_text_relocs = true;
581         } else {
582           ERROR("%s: Text relocations are not supported", GetPath());
583           return false;
584         }
585         break;

http://mxr.mozilla.org/mozilla-central/source/mozglue/linker/CustomElf.cpp#658
653            /* Treat as a DT_TEXTREL tag */
654            if (flags & DF_TEXTREL) {
655              if (strcmp("libflashplayer.so", GetName()) == 0) {
656                has_text_relocs = true;
657              } else {
658                ERROR("%s: Text relocations are not supported", GetPath());
659                return false;
660              }
661            }

Mike:
1) Should this have resulted in the run being marked as a failure? I'm presuming not.
2) If not, should this be marked as a warning, rather than an error?

Thanks :-)

Comment 1

[Mike Hommey [:glandium]]

(In reply to (Away 12th-20th July) [:edmorley] from comment #0)
> 1) Should this have resulted in the run being marked as a failure? I'm
> presuming not.
> 2) If not, should this be marked as a warning, rather than an error?

Mmmmm so it's a tough one. Here's what happens:
- custom linker tries to load library and fails.
- fallbacks to loading the library with the system linker and that works.

Now, the problem is that the latter can fail, at which point the errors you get from the custom linker are relevant. But if they were warnings, then you'd get them mixed with actual warnings, making it harder to know which is actually relevant to the failure...

Comment 2

[Ed Morley [:emorley]]

Ah I see, thank you