Parsing error in array: character array type mismatch; expected "[3 x i8]", got "[2 x i8]"

[jeshan]

I've got this error on one of my files. I've been able to narrow it down a bit.

Full gist: https://gist.github.com/jeshan/b8eb3cc7d7c12babedb1884b39b5e5b1

This gist was obtained when I compiled musl libc 1.2.2. In particular, its towctrans.c file produced the above assembly.

If you comment the module summary at the bottom (since #43 has not been implemented yet), you'll see the error:

character array type mismatch; expected "[3 x i8]", got "[2 x i8]"

This (terrible) string replace workaround on the @exceptions element would allow the parsing to succeed:

content = strings.Replace(content,
		"[2 x i8] c\"\\8E\\\\\", [2 x i8] c\"\\8F",
		"[3 x i8] c\"\\8E\\\\\", [3 x i8] c\"\\8F", 1)

So it looks like there's a bug handling the array elements at indices 8E and 8F (But I actually don't have a clue about what's going on yet)

Update: the exceptions array comes from this location. The code looks to be dealing with 2-element arrays but somehow we were expecting 3-element ones

[mewmew]

Thanks @jeshan for reporting this issue! You helped us uncover a bug in llir/llvm (which is now fixed in rev f16bc2c).

The issue was in enc.Unquote and was triggered by this code:

if b == '\\' && i+2 < len(s) {

where the i+2 was expecting two hexadecimal characters following the slash, but only one character may follow if it is the backslash-backslash escape sequence (\\). So, when a quoted string ends with a backslash (e.g. "\x8E\\"), the Unescape code was faulty.

Once more. Thanks for uncovering this issue, and for providing such detailed bug reports! They are invaluable when trouble-shooting issues like this. Kudos for posting the LLVM IR code as a Gist to help us reproduce the issue directly.

Cheerful regards,
Robin

[mewmew]

P.S. to strip the module summary aspects of the code example you posted, I ran the following LLVM command:

$ opt --verify -S -o foo.ll towctrans.lo.ll

--- towctrans.lo.ll	2021-05-28 00:13:15.316511389 +0200
+++ foo.ll	2021-05-28 00:42:46.446491703 +0200
@@ -1,4 +1,4 @@
-; ModuleID = 'towctrans.lo'
+; ModuleID = 'towctrans.lo.ll'
 source_filename = "src/ctype/towctrans.c"
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-alpine-linux-musl"
@@ -533,18 +533,3 @@
 !233 = !DILocation(line: 80, column: 18, scope: !228)
 !234 = !DILocation(line: 80, column: 9, scope: !228)
 !235 = !DILocation(line: 80, column: 2, scope: !228)
-
-^0 = module: (path: "towctrans.lo", hash: (0, 0, 0, 0, 0))
-^1 = gv: (name: "exceptions", summaries: (variable: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1)))) ; guid = 86732059878463693
-^2 = gv: (name: "casemap", summaries: (function: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 177, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0), refs: (^3, ^9, ^7, ^6, ^1)))) ; guid = 1801562777319867403
-^3 = gv: (name: "tab", summaries: (variable: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1)))) ; guid = 3242191788761116127
-^4 = gv: (name: "towlower_l", summaries: (alias: (module: ^0, flags: (linkage: weak, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), aliasee: ^10))) ; guid = 4742645945087594078
-^5 = gv: (name: "towlower", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0), calls: ((callee: ^2))))) ; guid = 5586237075964028068
-^6 = gv: (name: "rules", summaries: (variable: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1)))) ; guid = 8267841287748741871
-^7 = gv: (name: "rulebases", summaries: (variable: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1)))) ; guid = 10601644964188282356
-^8 = gv: (name: "__towupper_l", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 7, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0), calls: ((callee: ^13))))) ; guid = 10994762263732014571
-^9 = gv: (name: "casemap.mt", summaries: (variable: (module: ^0, flags: (linkage: internal, notEligibleToImport: 1, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1)))) ; guid = 11594124509780799275
-^10 = gv: (name: "__towlower_l", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 7, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0), calls: ((callee: ^5))))) ; guid = 12394728591314337204
-^11 = gv: (name: "llvm.dbg.declare") ; guid = 13513223491971101989
-^12 = gv: (name: "towupper_l", summaries: (alias: (module: ^0, flags: (linkage: weak, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), aliasee: ^8))) ; guid = 13980957050624971669
-^13 = gv: (name: "towupper", summaries: (function: (module: ^0, flags: (linkage: external, notEligibleToImport: 1, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0), calls: ((callee: ^2))))) ; guid = 14625676365288684025

[jeshan]

You helped us uncover a bug

Wow thanks for such a speedy fix. I was struggling with this for days so I should have reported sooner!

to strip the module summary aspects...

Good to know, thank you.

Happy summer. I hear it's a short and anticipated one in Sweden! 😄

[mewmew]

Happy summer. I hear it's a short and anticipated one in Sweden! smile

Most definitely! We have a saying in Sweden, Summer is the best day of the year ;)