1606012 - jit directive argument is null (build error)

Status: RESOLVED INVALID

(SeaMonkey :: General, defect)

Description

[democritus7]

User Agent: Mozilla/5.0 (X11; Linux i686; rv:60.0) Gecko/20100101 Firefox/60.0 SeaMonkey/2.53.1

Steps to reproduce:

Unzipped Bill's unofficial tarball, configured with --disable-warnings-as-errors, ran make.

Actual results:

Build process ended with this as the first of several error messages.
In member function ‘void js::jit::X86Encoding::BaseAssembler::movl_mr(int32_t, js::jit::X86Encoding::RegisterID, js::jit::X86Encoding::RegisterID, int, js::jit::X86Encoding::RegisterID)’,
inlined from ‘void js::jit::AssemblerX86Shared::movl.constprop(const js::jit::Operand&, js::jit::Register)’ at /home/me/Downloads/mozilla/buildsm/mozilla/js/src/jit/x86-shared/Assembler-x86-shared.h:494:25:
/home/me/Downloads/mozilla/buildsm/mozilla/js/src/jit/x86-shared/BaseAssembler-x86-shared.h:2166:13: error: ‘%s’ directive argument is null [-Werror=format-overflow=]
2166 | spew("movl " MEM_obs ", %s", ADDR_obs(offset, base, index, scale), GPReg32Name(dst));

There were several subsequent and related errors.

Expected results:

I would have expected that the build process would not incur an error, but since it did in the jit module, hopefully someone can examine it.

Comment 1

[Frank-Rainer Grahl (:frg)]

What compiler?

Comment 2

[democritus7]

I can't guarantee it, but am quite sure my run of configure resulted in g++ being the principal compiler. I deleted the source and build files before I saw this posted. My gcc package is 9.2.0. I have been looking for Bill G's source tarball of 2.53.1, but haven't yet found it.

Comment 3

[Frank-Rainer Grahl (:frg)]

gcc 9 is not yet supported. Recommended are gcc 6 to 8 or clang 7 or 8. In due time we will add support for gcc 9.

Bug 1514781 Comment 1 has some compiler options which might work or not. Looking at the error -Wno-error=format-overflow might fix it.

Bill is building directly from hg. I am not sure he has a public tarbally available.

Comment 4

[democritus7]

There was a source tarball indicated to be linked from this page https://www.seamonkey-project.org/releases/2.53.1b1#source, but it's been removed.

It's unclear whether the "null directive argument" stems from 32-bit gcc 9.2.0 or would also be issued by 32-bit gcc 8.

But chiefly, I would like to see if the repeated crashes I'm having with recent seamonkey has anything to do with the clang compiler, or if gcc builds result in similarly crash-prone 32-bit code. The high incidence of crashes makes browsing most unpleasant. It's not a browser I could recommend for serious browsing and email. Earlier seamonkeys never used to crash like this. There was a period around version 2.30 when seamonkey had a lot of crashes, but this is worse.