Using a boot prompt without a menu makes Syslinux 6.03-6 booting the default label only, ignoring the label that was entered. Auto completion of the defined labels works, though. Can anyone confirm this? Further, removing the DEFAULT <label> definition makes Syslinux panic, ignoring any boot label or command entered. Instead, without a default, I'd expect it to wait for user input, thus ignoring the TIMEOUT variable. Greetings from Berlin Alexander Freudenberg
On Sun, May 1, 2016 at 5:58 AM, Alexander Freudenberg via Syslinux <syslinux at zytor.com> wrote:> Using a boot prompt without a menu makes Syslinux 6.03-6 booting the defaultThe version appears to be a distribution-specific number.> label only, ignoring the label that was entered. Auto completion of theUsing the official builds (documented at http://www.syslinux.org/wiki/index.php/Download and available at https://www.kernel.org/pub/linux/utils/boot/syslinux/ ) and my personal builds, I've never seen this.> defined labels works, though. Can anyone confirm this?Could you post your entire config? If you have a short and simple config with ~2-3 LABELs that demonstrates your observations, that'd be excellent.> Further, removing the DEFAULT <label> definition makes Syslinux panic,Panic? Do you really mean "refuse to boot anything"> ignoring any boot label or command entered. Instead, without a default, I'd > expect it to wait for user input, thus ignoring the TIMEOUT variable.Quite odd but makes me think it's a distribution-specific issue or consequence of your configuration that I've never seen. -- -Gene
And apparently the original poster had bad headers. Delivery to the following recipient failed permanently: ml.afb at mailbox.org Technical details of permanent failure: Google tried to deliver your message, but it was rejected by the server for the recipient domain mailbox.org by mx2.mailbox.org. [80.241.60.215]. The error that the other server returned was: 577 5.1.1 <ml.afb at mailbox.org>: Recipient address rejected: undeliverable address: host imap.mailbox.org[80.241.60.199] said: 550 5.1.1 <ml.afb at mailbox.org> User doesn't exist: ml.afb at mailbox.org (in reply to RCPT TO command) On Sun, May 1, 2016 at 6:55 AM, Gene Cumm <gene.cumm at gmail.com> wrote:> On Sun, May 1, 2016 at 5:58 AM, Alexander Freudenberg via Syslinux > <syslinux at zytor.com> wrote: >> Using a boot prompt without a menu makes Syslinux 6.03-6 booting the default > > The version appears to be a distribution-specific number. > >> label only, ignoring the label that was entered. Auto completion of the > > Using the official builds (documented at > http://www.syslinux.org/wiki/index.php/Download and available at > https://www.kernel.org/pub/linux/utils/boot/syslinux/ ) and my > personal builds, I've never seen this. > >> defined labels works, though. Can anyone confirm this? > > Could you post your entire config? If you have a short and simple > config with ~2-3 LABELs that demonstrates your observations, that'd be > excellent. > >> Further, removing the DEFAULT <label> definition makes Syslinux panic, > > Panic? Do you really mean "refuse to boot anything" > >> ignoring any boot label or command entered. Instead, without a default, I'd >> expect it to wait for user input, thus ignoring the TIMEOUT variable. > > Quite odd but makes me think it's a distribution-specific issue or > consequence of your configuration that I've never seen. > > -- > -Gene-- -Gene
I'd suggest fixing your mail. ml.afb at mailbox.org still doesn't work. -- -Gene On Sun, May 1, 2016 at 6:55 AM, Gene Cumm <gene.cumm at gmail.com> wrote:> On Sun, May 1, 2016 at 5:58 AM, Alexander Freudenberg via Syslinux > <syslinux at zytor.com> wrote: >> Using a boot prompt without a menu makes Syslinux 6.03-6 booting the default > > The version appears to be a distribution-specific number. > >> label only, ignoring the label that was entered. Auto completion of the > > Using the official builds (documented at > http://www.syslinux.org/wiki/index.php/Download and available at > https://www.kernel.org/pub/linux/utils/boot/syslinux/ ) and my > personal builds, I've never seen this. > >> defined labels works, though. Can anyone confirm this? > > Could you post your entire config? If you have a short and simple > config with ~2-3 LABELs that demonstrates your observations, that'd be > excellent. > >> Further, removing the DEFAULT <label> definition makes Syslinux panic, > > Panic? Do you really mean "refuse to boot anything" > >> ignoring any boot label or command entered. Instead, without a default, I'd >> expect it to wait for user input, thus ignoring the TIMEOUT variable. > > Quite odd but makes me think it's a distribution-specific issue or > consequence of your configuration that I've never seen.
> Using a boot prompt without a menu makes Syslinux 6.03-6 booting the > default label only, ignoring the label that was entered. Auto completion > of the defined labels works, though. Can anyone confirm this? >I compiled binaries from the (hopefully, temporarily) "revert-9acbffd" branch from GeneC. The building environment is using old dependencies (i.e. not recent versions). With the resulting binaries, the aforementioned strange behavior was _not_ replicated -- _please_, do not just take this as valid / true; this same test _must_ be performed by others so to corroborate this result. But, using binaries from Fedora, 6.03-6.fc24, the problematic behavior _can_ be replicated. It has also been replicated by at least one additional user of ArchLinux, by using binaries from 6.03-6 (which is probably the package that was used by the original reporter). Additionally, at least one Debian Sid user is reporting the same unwanted behavior with 6.03+dfsg-12. Although the problem is not seen with 6.03+dfsg-11, I would not suggest that Debian should revert back. but rather that (additional) patches for gcc5+ compatibility should be also applied to Debian's packages. The abnormal behavior cannot be replicated with official 6.03 pre-built binaries downloaded from kernel.org. I have not bothered to test with 6.04-pre1 (as I would consider its binaries a waste of time, after they have failed me already in other tests). I should add that typing-in either labels or any other command, the result is the same: only / always the default command (label) is executed when pressing ENTER. To be accurate, I have not actually tested whether it is really the DEFAULT command, or the first LABEL, that is being executed. So, this seems to be some environmental / building issue -- again, please do not take this statement as valid / true unless these tests are replicated by others. Considering that most Linux distributions that are already using some package based on Syslinux 6.03 are also using relatively-updated versions of its dependencies (e.g. gcc5+, binutils 2.23+ and so on), I am wondering whether some of the pending problems related to makefiles and/or ld files could be the source of this unwanted behavior. For instance, "HIDDEN was added to ld in binutils version 2.23", and the current Syslinux source code (both, 6.03 and current git master head) is not using the necessary code to cope with this "HIDDEN". Perhaps it is time to add some kind of conditional code in Syslinux's "*.ld" files , depending on the version of binutils, so binaries could be functional in both cases (binutils prior and/or post/equal to version 2.23). Of course this is just an example, and it might not be the culprit of this problem. Yet, the basic concept might be relevant for whatever might be generating this issue. Another possible issue that might come to mind: building with/without "CFLAGS += -fno-PIE" or something of that sort. These are wild uneducated guesses; so much so that calling it a "guess" might already be too much. I hope knowledgeable developers will find a solution. Regards, Ady.
On Sat, May 07, 2016 at 09:56:59PM +0300, Ady via Syslinux wrote:> On Sun, May 01, 2016 at 11:58:41AM +0200, Alexander Freudenberg via Syslinux wrote: > > Using a boot prompt without a menu makes Syslinux 6.03-6 booting the > > default label only, ignoring the label that was entered. Auto completion > > of the defined labels works, though. Can anyone confirm this? > > > > I compiled binaries from the (hopefully, temporarily) "revert-9acbffd" > branch from GeneC. The building environment is using old dependencies > (i.e. not recent versions). > > With the resulting binaries, the aforementioned strange behavior was > _not_ replicated -- _please_, do not just take this as valid / true; > this same test _must_ be performed by others so to corroborate this > result. > > But, using binaries from Fedora, 6.03-6.fc24, the problematic behavior > _can_ be replicated. > > It has also been replicated by at least one additional user of > ArchLinux, by using binaries from 6.03-6 (which is probably the package > that was used by the original reporter). > > Additionally, at least one Debian Sid user is reporting the same > unwanted behavior with 6.03+dfsg-12. Although the problem is not seen > with 6.03+dfsg-11, I would not suggest that Debian should revert back. > but rather that (additional) patches for gcc5+ compatibility should be > also applied to Debian's packages. > > The abnormal behavior cannot be replicated with official 6.03 pre-built > binaries downloaded from kernel.org. > > I have not bothered to test with 6.04-pre1 (as I would consider its > binaries a waste of time, after they have failed me already in other > tests). > > I should add that typing-in either labels or any other command, the > result is the same: only / always the default command (label) is > executed when pressing ENTER. To be accurate, I have not actually > tested whether it is really the DEFAULT command, or the first LABEL, > that is being executed. > > So, this seems to be some environmental / building issue -- again, > please do not take this statement as valid / true unless these tests > are replicated by others. > > Considering that most Linux distributions that are already using some > package based on Syslinux 6.03 are also using relatively-updated > versions of its dependencies (e.g. gcc5+, binutils 2.23+ and so on), I > am wondering whether some of the pending problems related to makefiles > and/or ld files could be the source of this unwanted behavior. > > For instance, "HIDDEN was added to ld in binutils version 2.23", and > the current Syslinux source code (both, 6.03 and current git master > head) is not using the necessary code to cope with this "HIDDEN". > Perhaps it is time to add some kind of conditional code in Syslinux's > "*.ld" files , depending on the version of binutils, so binaries could > be functional in both cases (binutils prior and/or post/equal to > version 2.23). > > Of course this is just an example, and it might not be the culprit of > this problem. Yet, the basic concept might be relevant for whatever > might be generating this issue. > > Another possible issue that might come to mind: building with/without > "CFLAGS += -fno-PIE" or something of that sort. > > These are wild uneducated guesses; so much so that calling it a "guess" > might already be too much. I hope knowledgeable developers will find a > solution. > > Regards, > Ady. >Turns out, the undesired behaviour must be somewhat specific to the build in Arch Linux' repository. After building Syslinux from the official source, it boots any label entered during boot prompt correctly. (I have not tested removing the DEFAULT statement, though) Thank you for your help - Alex